Analysis log for: Selective Dry Cow Therapy can be Implemented Successfully in Cows of All Milk Production Levels
Sam Rowe (samrowe101@gmail.com)
April 2022
- Acknowledgements
- Inspect Data
- Analysis
- DHI Dataset
- Milk yield at DO —
- Repeat analysis with different definition for my3 (i.e within herd)
- Analysis
- DHI Dataset
- Milk yield at DO —
library(knitr)
library(readr)
library(lme4)
library(emmeans)
library(geepack)
library(table1)
library(ggplot2)
library(tidyverse)
library(janitor)
library(psych)
library(performance)
cow <- readRDS("~/Dropbox/R backup/2021/SDCT.Milkyield/sdct.my.cow.rds")
qtr <- readRDS("~/Dropbox/R backup/2021/SDCT.Milkyield/sdct.my.qtr.rds")
dhi <- readRDS("~/Dropbox/R backup/2021/SDCT.Milkyield/sdct.my.dhi.rds")
Acknowledgements
Sam Rowe*1,2 Fidele Kabera,3,4 Simon Dufour,3,4 Sandra Godden,2 Jean-Philippe Roy4,5 Daryl Nydam6
1Sydney School of Veterinary Science, The University of Sydney, Camden, New South Wales 2570, Australia 2Department of Veterinary Population Medicine, University of Minnesota, St. Paul 55108 3Département de pathologie et microbiologie, Faculté de médecine vétérinaire, Université de Montréal, Saint-Hyacinthe, Québec, Canada J2S 2M2 4Mastitis Network, Saint-Hyacinthe, Québec, Canada J2S 2M2 5Département de sciences cliniques, Faculté de médecine vétérinaire, Université de Montréal, Saint-Hyacinthe, Québec, Canada J2S 2M2 6Department of Public and Ecosystem Health, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853, USA;
Inspect Data
qtr <- qtr %>% ungroup()
dhi <- dhi %>% ungroup()
cow <- cow %>% ungroup()
#summarytools::dfSummary(Baseline, style='grid')
print(summarytools::dfSummary(qtr,valid.col=FALSE, graph.magnif=0.8, style="grid"), method = "render")Data Frame Summary
qtr
Dimensions: 5097 x 23Duplicates: 0
| No | Variable | Stats / Values | Freqs (% of Valid) | Graph | Missing | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | FARMID [numeric] |
|
16 distinct values |  |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 | CowID [numeric] |
|
1265 distinct values |  |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 | QTR [character] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 4 | ab [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 5 | Parity [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 6 | CalvDat [POSIXct, POSIXt] |
|
266 distinct values |  |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 7 | DOSCC [numeric] |
|
429 distinct values |  |
4 (0.1%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 8 | DOMY [numeric] |
|
238 distinct values |  |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 9 | Enrolldat [POSIXct, POSIXt] |
|
136 distinct values |  |
1 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 10 | Tx [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 11 | DOCult1 [character] |
|
|
 |
409 (8.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 12 | DOCult2 [character] |
|
|
 |
5012 (98.3%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 13 | PCCult1 [character] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 14 | PCCult2 [character] |
|
|
 |
4978 (97.7%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 15 | source [character] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 16 | cure [numeric] |
|
|
 |
3910 (76.7%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 17 | newimi [numeric] |
|
|
 |
459 (9.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 18 | imipc [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 19 | my3 [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 20 | my3h [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 21 | imido [numeric] |
|
|
 |
409 (8.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 22 | alg [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 23 | cult [numeric] |
|
|
 |
0 (0.0%) |
Generated by summarytools 1.0.1 (R version 4.2.1)
2022-09-26
print(summarytools::dfSummary(dhi,valid.col=FALSE, graph.magnif=0.8, style="grid"), method = "render")Data Frame Summary
dhi
Dimensions: 4553 x 19Duplicates: 0
| No | Variable | Stats / Values | Freqs (% of Valid) | Graph | Missing | ||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | CowID [numeric] |
|
1283 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 2 | FARMID [numeric] |
|
16 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 3 | Parity [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 4 | CalvDat [Date] |
|
242 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 5 | DOSCC [numeric] |
|
424 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 6 | DOMY [numeric] |
|
215 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 7 | Enrolldat [Date] |
|
115 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 8 | Tx [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 9 | TestNo [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 10 | DIM [numeric] |
|
120 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 11 | MY [numeric] |
|
399 distinct values |  |
2 (0.0%) | ||||||||||||||||||||||||||||||
| 12 | SCC [numeric] |
|
665 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 13 | TestDIMcat20 [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 14 | my3 [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 15 | my3h [factor] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 16 | scc [numeric] |
|
665 distinct values |  |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 17 | source [character] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 18 | alg [numeric] |
|
|
 |
0 (0.0%) | ||||||||||||||||||||||||||||||
| 19 | cult [numeric] |
|
|
 |
0 (0.0%) |
Generated by summarytools 1.0.1 (R version 4.2.1)
2022-09-26
print(summarytools::dfSummary(cow,valid.col=FALSE, graph.magnif=0.8, style="grid"), method = "render")Data Frame Summary
cow
Dimensions: 1485 x 19Duplicates: 0
| No | Variable | Stats / Values | Freqs (% of Valid) | Graph | Missing | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | CowID [numeric] |
|
1388 distinct values |  |
0 (0.0%) | |||||||||||||||
| 2 | FARMID [numeric] |
|
16 distinct values |  |
0 (0.0%) | |||||||||||||||
| 3 | Parity [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 4 | Enrolldat [POSIXct, POSIXt] |
|
143 distinct values |  |
0 (0.0%) | |||||||||||||||
| 5 | Tx [factor] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 6 | CalvDat [Date] |
|
287 distinct values |  |
0 (0.0%) | |||||||||||||||
| 7 | DOMY [numeric] |
|
248 distinct values |  |
0 (0.0%) | |||||||||||||||
| 8 | DOSCC [numeric] |
|
454 distinct values |  |
1 (0.1%) | |||||||||||||||
| 9 | CM [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 10 | CMTAR [numeric] |
|
109 distinct values |  |
0 (0.0%) | |||||||||||||||
| 11 | CMDP [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 12 | CullDP [numeric] | 1 distinct value |
|
 |
0 (0.0%) | |||||||||||||||
| 13 | Cull2TAR [numeric] |
|
80 distinct values |  |
0 (0.0%) | |||||||||||||||
| 14 | Cull2 [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 15 | my3 [factor] |
|
|
|
0 (0.0%) | |||||||||||||||
| 16 | my3h [factor] |
|
|
|
0 (0.0%) | |||||||||||||||
| 17 | source [character] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 18 | alg [numeric] |
|
|
 |
0 (0.0%) | |||||||||||||||
| 19 | cult [numeric] |
|
|
|
0 (0.0%) |
Generated by summarytools 1.0.1 (R version 4.2.1)
2022-09-26
Report milk culture results at dry-off
do <- table(c(qtr$DOCult1,qtr$DOCult2)) %>% as.data.frame
do$atrisk <- nrow(qtr[!is.na(qtr$imido),])
do$percent <- round((do$Freq / do$atrisk) * 100,2)
do %>% select(Var1,Freq,percent)## Var1 Freq percent
## 1 Acinetobacter guillouiae 1 0.02
## 2 Acinetobacter sp. 5 0.11
## 3 Actinomyces sp. 2 0.04
## 4 Aerococcus sp. 14 0.30
## 5 Aerococcus viridans 13 0.28
## 6 Bacillus sp. 37 0.79
## 7 Brevibacillus agri 1 0.02
## 8 Brevundimonas sp. 1 0.02
## 9 Chryseobacterium sp. 1 0.02
## 10 Corynebacterium sp. 177 3.78
## 11 Enterobacter aerogenes 1 0.02
## 12 Enterobacter cloacae 9 0.19
## 13 Enterobacter kobei 1 0.02
## 14 Enterococcus faecalis 1 0.02
## 15 Enterococcus faecium 1 0.02
## 16 Enterococcus hirae 2 0.04
## 17 Enterococcus sp. 6 0.13
## 18 Escherichia coli 7 0.15
## 19 Escherichia sp. 3 0.06
## 20 Gram negative organism 6 0.13
## 21 Gram positive coccus 11 0.23
## 22 Gram positive organism 14 0.30
## 23 Gram positive rod 21 0.45
## 24 Klebsiella pneumoniae 4 0.09
## 25 Kocuria sp. 5 0.11
## 26 Kocuria species 3 0.06
## 27 Lactobacillus sp. 3 0.06
## 28 Lactococcus garvieae 8 0.17
## 29 Lactococcus lactis 5 0.11
## 30 Lactococcus sp. 11 0.23
## 31 Micrococcus luteus 3 0.06
## 32 Micrococcus sp. 55 1.17
## 33 Neisseria sp. 4 0.09
## 34 No Growth 3501 74.68
## 35 Paenibacillus sp. 3 0.06
## 36 Pseudomonas aeruginosa 3 0.06
## 37 Pseudomonas sp. 11 0.23
## 38 Rothia sp. 6 0.13
## 39 Staphylococcus aureus 26 0.55
## 40 Staphylococcus auricularis 1 0.02
## 41 Staphylococcus capitis 6 0.13
## 42 Staphylococcus chromogenes 329 7.02
## 43 Staphylococcus cohnii 3 0.06
## 44 Staphylococcus epidermidis 36 0.77
## 45 Staphylococcus equorum 11 0.23
## 46 Staphylococcus gallinarum 1 0.02
## 47 Staphylococcus haemolyticus 47 1.00
## 48 Staphylococcus hominis 5 0.11
## 49 Staphylococcus sciuri 19 0.41
## 50 Staphylococcus simulans 11 0.23
## 51 Staphylococcus sp. 200 4.27
## 52 Staphylococcus succinus 1 0.02
## 53 Staphylococcus xylosus/saprophyticus 60 1.28
## 54 Stenotrophomonas sp. 1 0.02
## 55 Streptococcus mitis 6 0.13
## 56 Streptococcus oralis 2 0.04
## 57 Streptococcus sp. 33 0.70
## 58 Streptococcus uberis 8 0.17
## 59 Streptomyces sp. 2 0.04
## 60 Trueperella pyogenes 2 0.04
## 61 Yeast 3 0.06Report milk culture results at calving
pc <- table(c(qtr$PCCult1,qtr$PCCult2)) %>% as.data.frame
pc$atrisk <- nrow(qtr[!is.na(qtr$imipc),])
pc$percent <- round((pc$Freq / pc$atrisk) * 100,2)
pc %>% select(Var1,Freq,percent)## Var1 Freq percent
## 1 Acinetobacter sp. 19 0.37
## 2 Aerococcus sp. 10 0.20
## 3 Aerococcus viridans 19 0.37
## 4 Arthrobacter sp. 2 0.04
## 5 Bacillus sp. 131 2.57
## 6 Brachybacterium faecium 1 0.02
## 7 Brachybacterium sp. 2 0.04
## 8 Brevundimonas sp. 2 0.04
## 9 Chryseobacterium sp. 1 0.02
## 10 Corynebacterium sp. 131 2.57
## 11 Enterobacter cloacae 5 0.10
## 12 Enterococcus faecalis 4 0.08
## 13 Enterococcus hirae 1 0.02
## 14 Enterococcus saccharolyticus 3 0.06
## 15 Enterococcus sp. 3 0.06
## 16 Enterococcus thailandicus 1 0.02
## 17 Escherichia coli 9 0.18
## 18 Escherichia sp. 2 0.04
## 19 Gram negative organism 19 0.37
## 20 Gram positive coccus 18 0.35
## 21 Gram positive organism 15 0.29
## 22 Gram positive rod 44 0.86
## 23 Klebsiella oxytoca 3 0.06
## 24 Klebsiella pneumoniae 1 0.02
## 25 Kocuria carniphila 1 0.02
## 26 Kocuria palustris 1 0.02
## 27 Kocuria sp. 1 0.02
## 28 Lactobacillus sp. 1 0.02
## 29 Lactococcus garvieae 6 0.12
## 30 Lactococcus lactis 1 0.02
## 31 Lactococcus sp. 6 0.12
## 32 Micrococcus luteus 6 0.12
## 33 Micrococcus sp. 8 0.16
## 34 Moraxella sp. 2 0.04
## 35 Neisseria flavescens 2 0.04
## 36 Neisseria sp. 4 0.08
## 37 No Growth 3977 78.03
## 38 Oligella sp. 1 0.02
## 39 Paenibacillus sp. 2 0.04
## 40 Pantoea sp. 6 0.12
## 41 Prototheca sp. 8 0.16
## 42 Pseudomonas aeruginosa 1 0.02
## 43 Pseudomonas sp. 4 0.08
## 44 Rothia sp. 20 0.39
## 45 Serratia sp. 1 0.02
## 46 Staphylococcus aureus 18 0.35
## 47 Staphylococcus capitis 3 0.06
## 48 Staphylococcus chromogenes 198 3.88
## 49 Staphylococcus cohnii 1 0.02
## 50 Staphylococcus epidermidis 4 0.08
## 51 Staphylococcus equorum 1 0.02
## 52 Staphylococcus gallinarum 1 0.02
## 53 Staphylococcus haemolyticus 38 0.75
## 54 Staphylococcus hominis 4 0.08
## 55 Staphylococcus sciuri 136 2.67
## 56 Staphylococcus simulans 5 0.10
## 57 Staphylococcus sp. 220 4.32
## 58 Staphylococcus xylosus/saprophyticus 30 0.59
## 59 Stenotrophomonas maltophilia 5 0.10
## 60 Streptococcus dysgalactiae 4 0.08
## 61 Streptococcus gallinarum 1 0.02
## 62 Streptococcus mitis 7 0.14
## 63 Streptococcus sp. 21 0.41
## 64 Streptococcus uberis 3 0.06
## 65 Trueperella pyogenes 2 0.04
## 66 Weissella confusa 1 0.02
## 67 Yeast 8 0.16table1(~ my3 + factor(ab) + factor(imido) + DOMY + DOSCC + factor(Parity) + factor(FARMID) + source | Tx, data=qtr, na.is.category = F)| C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
Overall (N=5097) |
|
|---|---|---|---|---|
| my3 | ||||
| Low production | 582 (31.5%) | 411 (30.3%) | 706 (37.3%) | 1699 (33.3%) |
| Mid production | 636 (34.4%) | 442 (32.6%) | 621 (32.8%) | 1699 (33.3%) |
| High production | 631 (34.1%) | 502 (37.0%) | 566 (29.9%) | 1699 (33.3%) |
| factor(ab) | ||||
| 0 | 1049 (56.7%) | 759 (56.0%) | 0 (0%) | 1808 (35.5%) |
| 1 | 800 (43.3%) | 596 (44.0%) | 1893 (100%) | 3289 (64.5%) |
| factor(imido) | ||||
| 0 | 1274 (74.6%) | 938 (75.6%) | 1289 (74.1%) | 3501 (74.7%) |
| 1 | 434 (25.4%) | 303 (24.4%) | 450 (25.9%) | 1187 (25.3%) |
| Missing | 141 (7.6%) | 114 (8.4%) | 154 (8.1%) | 409 (8.0%) |
| DOMY | ||||
| Mean (SD) | 26.8 (8.56) | 27.1 (8.45) | 25.8 (8.61) | 26.5 (8.57) |
| Median [Min, Max] | 27.7 [1.81, 49.4] | 28.6 [2.72, 49.4] | 26.4 [4.54, 49.4] | 27.2 [1.81, 49.4] |
| DOSCC | ||||
| Mean (SD) | 4.36 (1.22) | 4.43 (1.20) | 4.48 (1.18) | 4.42 (1.20) |
| Median [Min, Max] | 4.20 [1.61, 8.59] | 4.38 [1.61, 7.59] | 4.40 [1.61, 8.35] | 4.33 [1.61, 8.59] |
| Missing | 0 (0%) | 0 (0%) | 4 (0.2%) | 4 (0.1%) |
| factor(Parity) | ||||
| 1 | 872 (47.2%) | 599 (44.2%) | 765 (40.4%) | 2236 (43.9%) |
| 2 | 472 (25.5%) | 428 (31.6%) | 572 (30.2%) | 1472 (28.9%) |
| 3 | 505 (27.3%) | 328 (24.2%) | 556 (29.4%) | 1389 (27.3%) |
| factor(FARMID) | ||||
| 1 | 100 (5.41%) | 83 (6.13%) | 112 (5.92%) | 295 (5.79%) |
| 2 | 92 (4.98%) | 124 (9.15%) | 108 (5.71%) | 324 (6.36%) |
| 3 | 291 (15.7%) | 259 (19.1%) | 285 (15.1%) | 835 (16.4%) |
| 4 | 456 (24.7%) | 424 (31.3%) | 390 (20.6%) | 1270 (24.9%) |
| 5 | 310 (16.8%) | 278 (20.5%) | 299 (15.8%) | 887 (17.4%) |
| 6 | 61 (3.30%) | 55 (4.06%) | 52 (2.75%) | 168 (3.30%) |
| 7 | 116 (6.27%) | 132 (9.74%) | 146 (7.71%) | 394 (7.73%) |
| 8 | 32 (1.73%) | 0 (0%) | 36 (1.90%) | 68 (1.33%) |
| 9 | 32 (1.73%) | 0 (0%) | 44 (2.32%) | 76 (1.49%) |
| 10 | 32 (1.73%) | 0 (0%) | 49 (2.59%) | 81 (1.59%) |
| 11 | 43 (2.33%) | 0 (0%) | 35 (1.85%) | 78 (1.53%) |
| 12 | 28 (1.51%) | 0 (0%) | 40 (2.11%) | 68 (1.33%) |
| 13 | 40 (2.16%) | 0 (0%) | 54 (2.85%) | 94 (1.84%) |
| 14 | 112 (6.06%) | 0 (0%) | 136 (7.18%) | 248 (4.87%) |
| 15 | 20 (1.08%) | 0 (0%) | 35 (1.85%) | 55 (1.08%) |
| 16 | 84 (4.54%) | 0 (0%) | 72 (3.80%) | 156 (3.06%) |
| source | ||||
| CA | 423 (22.9%) | 0 (0%) | 501 (26.5%) | 924 (18.1%) |
| US | 1426 (77.1%) | 1355 (100%) | 1392 (73.5%) | 4173 (81.9%) |
table1(~ DOMY + my3 + source + factor(Parity) + factor(FARMID) + DOSCC | Tx, data=cow, na.is.category = F)| C_SDCT (N=536) |
A_SDCT (N=394) |
BDCT (N=555) |
Overall (N=1485) |
|
|---|---|---|---|---|
| DOMY | ||||
| Mean (SD) | 26.9 (8.60) | 27.3 (8.38) | 26.0 (8.52) | 26.7 (8.52) |
| Median [Min, Max] | 27.7 [1.81, 49.4] | 29.0 [2.72, 49.4] | 27.0 [4.54, 49.4] | 27.7 [1.81, 49.4] |
| my3 | ||||
| Low production | 171 (31.9%) | 121 (30.7%) | 203 (36.6%) | 495 (33.3%) |
| Mid production | 182 (34.0%) | 127 (32.2%) | 186 (33.5%) | 495 (33.3%) |
| High production | 183 (34.1%) | 146 (37.1%) | 166 (29.9%) | 495 (33.3%) |
| source | ||||
| CA | 126 (23.5%) | 0 (0%) | 148 (26.7%) | 274 (18.5%) |
| US | 410 (76.5%) | 394 (100%) | 407 (73.3%) | 1211 (81.5%) |
| factor(Parity) | ||||
| 1 | 241 (45.0%) | 167 (42.4%) | 218 (39.3%) | 626 (42.2%) |
| 2 | 141 (26.3%) | 123 (31.2%) | 170 (30.6%) | 434 (29.2%) |
| 3 | 154 (28.7%) | 104 (26.4%) | 167 (30.1%) | 425 (28.6%) |
| factor(FARMID) | ||||
| 1 | 27 (5.04%) | 21 (5.33%) | 28 (5.05%) | 76 (5.12%) |
| 2 | 24 (4.48%) | 31 (7.87%) | 27 (4.86%) | 82 (5.52%) |
| 3 | 87 (16.2%) | 83 (21.1%) | 93 (16.8%) | 263 (17.7%) |
| 4 | 136 (25.4%) | 130 (33.0%) | 122 (22.0%) | 388 (26.1%) |
| 5 | 84 (15.7%) | 77 (19.5%) | 81 (14.6%) | 242 (16.3%) |
| 6 | 16 (2.99%) | 15 (3.81%) | 16 (2.88%) | 47 (3.16%) |
| 7 | 36 (6.72%) | 37 (9.39%) | 40 (7.21%) | 113 (7.61%) |
| 8 | 9 (1.68%) | 0 (0%) | 9 (1.62%) | 18 (1.21%) |
| 9 | 12 (2.24%) | 0 (0%) | 13 (2.34%) | 25 (1.68%) |
| 10 | 9 (1.68%) | 0 (0%) | 14 (2.52%) | 23 (1.55%) |
| 11 | 11 (2.05%) | 0 (0%) | 11 (1.98%) | 22 (1.48%) |
| 12 | 8 (1.49%) | 0 (0%) | 11 (1.98%) | 19 (1.28%) |
| 13 | 14 (2.61%) | 0 (0%) | 15 (2.70%) | 29 (1.95%) |
| 14 | 34 (6.34%) | 0 (0%) | 42 (7.57%) | 76 (5.12%) |
| 15 | 6 (1.12%) | 0 (0%) | 10 (1.80%) | 16 (1.08%) |
| 16 | 23 (4.29%) | 0 (0%) | 23 (4.14%) | 46 (3.10%) |
| DOSCC | ||||
| Mean (SD) | 4.39 (1.23) | 4.45 (1.21) | 4.46 (1.18) | 4.43 (1.21) |
| Median [Min, Max] | 4.26 [1.61, 8.59] | 4.39 [1.61, 8.25] | 4.38 [1.61, 8.35] | 4.34 [1.61, 8.59] |
| Missing | 0 (0%) | 0 (0%) | 1 (0.2%) | 1 (0.1%) |
table1(~ my3 + factor(ab) + factor(imido) + factor(FARMID) | source, data=qtr, na.is.category = F)| CA (N=924) |
US (N=4173) |
Overall (N=5097) |
|
|---|---|---|---|
| my3 | |||
| Low production | 441 (47.7%) | 1258 (30.1%) | 1699 (33.3%) |
| Mid production | 322 (34.8%) | 1377 (33.0%) | 1699 (33.3%) |
| High production | 161 (17.4%) | 1538 (36.9%) | 1699 (33.3%) |
| factor(ab) | |||
| 0 | 243 (26.3%) | 1565 (37.5%) | 1808 (35.5%) |
| 1 | 681 (73.7%) | 2608 (62.5%) | 3289 (64.5%) |
| factor(imido) | |||
| 0 | 642 (71.8%) | 2859 (75.4%) | 3501 (74.7%) |
| 1 | 252 (28.2%) | 935 (24.6%) | 1187 (25.3%) |
| Missing | 30 (3.2%) | 379 (9.1%) | 409 (8.0%) |
| factor(FARMID) | |||
| 1 | 0 (0%) | 295 (7.07%) | 295 (5.79%) |
| 2 | 0 (0%) | 324 (7.76%) | 324 (6.36%) |
| 3 | 0 (0%) | 835 (20.0%) | 835 (16.4%) |
| 4 | 0 (0%) | 1270 (30.4%) | 1270 (24.9%) |
| 5 | 0 (0%) | 887 (21.3%) | 887 (17.4%) |
| 6 | 0 (0%) | 168 (4.03%) | 168 (3.30%) |
| 7 | 0 (0%) | 394 (9.44%) | 394 (7.73%) |
| 8 | 68 (7.36%) | 0 (0%) | 68 (1.33%) |
| 9 | 76 (8.23%) | 0 (0%) | 76 (1.49%) |
| 10 | 81 (8.77%) | 0 (0%) | 81 (1.59%) |
| 11 | 78 (8.44%) | 0 (0%) | 78 (1.53%) |
| 12 | 68 (7.36%) | 0 (0%) | 68 (1.33%) |
| 13 | 94 (10.2%) | 0 (0%) | 94 (1.84%) |
| 14 | 248 (26.8%) | 0 (0%) | 248 (4.87%) |
| 15 | 55 (5.95%) | 0 (0%) | 55 (1.08%) |
| 16 | 156 (16.9%) | 0 (0%) | 156 (3.06%) |
table1(~ my3 + DOMY + DOSCC + factor(Parity) + factor(FARMID) + Tx | source, data=cow, na.is.category = F)| CA (N=274) |
US (N=1211) |
Overall (N=1485) |
|
|---|---|---|---|
| my3 | |||
| Low production | 134 (48.9%) | 361 (29.8%) | 495 (33.3%) |
| Mid production | 96 (35.0%) | 399 (32.9%) | 495 (33.3%) |
| High production | 44 (16.1%) | 451 (37.2%) | 495 (33.3%) |
| DOMY | |||
| Mean (SD) | 23.7 (6.82) | 27.4 (8.73) | 26.7 (8.52) |
| Median [Min, Max] | 23.8 [6.40, 43.1] | 29.0 [1.81, 49.4] | 27.7 [1.81, 49.4] |
| DOSCC | |||
| Mean (SD) | 4.43 (1.16) | 4.43 (1.22) | 4.43 (1.21) |
| Median [Min, Max] | 4.37 [1.61, 7.69] | 4.33 [1.61, 8.59] | 4.34 [1.61, 8.59] |
| Missing | 1 (0.4%) | 0 (0%) | 1 (0.1%) |
| factor(Parity) | |||
| 1 | 116 (42.3%) | 510 (42.1%) | 626 (42.2%) |
| 2 | 70 (25.5%) | 364 (30.1%) | 434 (29.2%) |
| 3 | 88 (32.1%) | 337 (27.8%) | 425 (28.6%) |
| factor(FARMID) | |||
| 1 | 0 (0%) | 76 (6.28%) | 76 (5.12%) |
| 2 | 0 (0%) | 82 (6.77%) | 82 (5.52%) |
| 3 | 0 (0%) | 263 (21.7%) | 263 (17.7%) |
| 4 | 0 (0%) | 388 (32.0%) | 388 (26.1%) |
| 5 | 0 (0%) | 242 (20.0%) | 242 (16.3%) |
| 6 | 0 (0%) | 47 (3.88%) | 47 (3.16%) |
| 7 | 0 (0%) | 113 (9.33%) | 113 (7.61%) |
| 8 | 18 (6.57%) | 0 (0%) | 18 (1.21%) |
| 9 | 25 (9.12%) | 0 (0%) | 25 (1.68%) |
| 10 | 23 (8.39%) | 0 (0%) | 23 (1.55%) |
| 11 | 22 (8.03%) | 0 (0%) | 22 (1.48%) |
| 12 | 19 (6.93%) | 0 (0%) | 19 (1.28%) |
| 13 | 29 (10.6%) | 0 (0%) | 29 (1.95%) |
| 14 | 76 (27.7%) | 0 (0%) | 76 (5.12%) |
| 15 | 16 (5.84%) | 0 (0%) | 16 (1.08%) |
| 16 | 46 (16.8%) | 0 (0%) | 46 (3.10%) |
| Tx | |||
| C_SDCT | 126 (46.0%) | 410 (33.9%) | 536 (36.1%) |
| A_SDCT | 0 (0%) | 394 (32.5%) | 394 (26.5%) |
| BDCT | 148 (54.0%) | 407 (33.6%) | 555 (37.4%) |
table1(~ factor(ab) + factor(imido) + DOMY + DOSCC + factor(Parity) | my3*Tx, data=qtr, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=582) |
A_SDCT (N=411) |
BDCT (N=706) |
C_SDCT (N=636) |
A_SDCT (N=442) |
BDCT (N=621) |
C_SDCT (N=631) |
A_SDCT (N=502) |
BDCT (N=566) |
C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
|
| factor(ab) | ||||||||||||
| 0 | 345 (59.3%) | 162 (39.4%) | 0 (0%) | 371 (58.3%) | 271 (61.3%) | 0 (0%) | 333 (52.8%) | 326 (64.9%) | 0 (0%) | 1049 (56.7%) | 759 (56.0%) | 0 (0%) |
| 1 | 237 (40.7%) | 249 (60.6%) | 706 (100%) | 265 (41.7%) | 171 (38.7%) | 621 (100%) | 298 (47.2%) | 176 (35.1%) | 566 (100%) | 800 (43.3%) | 596 (44.0%) | 1893 (100%) |
| factor(imido) | ||||||||||||
| 0 | 436 (79.6%) | 309 (82.0%) | 496 (75.5%) | 439 (74.4%) | 304 (74.7%) | 438 (76.2%) | 399 (70.0%) | 325 (71.1%) | 355 (70.0%) | 1274 (74.6%) | 938 (75.6%) | 1289 (74.1%) |
| 1 | 112 (20.4%) | 68 (18.0%) | 161 (24.5%) | 151 (25.6%) | 103 (25.3%) | 137 (23.8%) | 171 (30.0%) | 132 (28.9%) | 152 (30.0%) | 434 (25.4%) | 303 (24.4%) | 450 (25.9%) |
| Missing | 34 (5.8%) | 34 (8.3%) | 49 (6.9%) | 46 (7.2%) | 35 (7.9%) | 46 (7.4%) | 61 (9.7%) | 45 (9.0%) | 59 (10.4%) | 141 (7.6%) | 114 (8.4%) | 154 (8.1%) |
| DOMY | ||||||||||||
| Mean (SD) | 16.7 (5.14) | 16.8 (5.77) | 16.9 (5.01) | 27.3 (2.11) | 27.7 (2.06) | 27.2 (2.00) | 35.5 (3.91) | 35.0 (3.44) | 35.5 (4.01) | 26.8 (8.56) | 27.1 (8.45) | 25.8 (8.61) |
| Median [Min, Max] | 17.6 [1.81, 23.6] | 18.6 [2.72, 23.6] | 18.1 [4.54, 23.6] | 27.4 [23.6, 30.4] | 27.7 [23.6, 30.4] | 27.2 [23.6, 30.4] | 34.5 [30.4, 49.4] | 34.5 [30.4, 49.4] | 34.5 [30.4, 49.4] | 27.7 [1.81, 49.4] | 28.6 [2.72, 49.4] | 26.4 [4.54, 49.4] |
| DOSCC | ||||||||||||
| Mean (SD) | 5.03 (1.30) | 4.93 (1.20) | 5.01 (1.20) | 4.19 (1.06) | 4.21 (1.12) | 4.24 (1.08) | 3.92 (1.03) | 4.22 (1.15) | 4.08 (1.02) | 4.36 (1.22) | 4.43 (1.20) | 4.48 (1.18) |
| Median [Min, Max] | 5.02 [1.95, 8.17] | 4.91 [1.61, 7.35] | 4.88 [2.60, 8.35] | 4.11 [1.95, 7.89] | 4.25 [1.79, 7.56] | 4.19 [1.61, 7.38] | 3.85 [1.61, 8.59] | 4.13 [2.30, 7.59] | 3.99 [1.61, 6.61] | 4.20 [1.61, 8.59] | 4.38 [1.61, 7.59] | 4.40 [1.61, 8.35] |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 4 (0.6%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 4 (0.2%) |
| factor(Parity) | ||||||||||||
| 1 | 223 (38.3%) | 132 (32.1%) | 259 (36.7%) | 291 (45.8%) | 202 (45.7%) | 259 (41.7%) | 358 (56.7%) | 265 (52.8%) | 247 (43.6%) | 872 (47.2%) | 599 (44.2%) | 765 (40.4%) |
| 2 | 155 (26.6%) | 158 (38.4%) | 226 (32.0%) | 168 (26.4%) | 133 (30.1%) | 176 (28.3%) | 149 (23.6%) | 137 (27.3%) | 170 (30.0%) | 472 (25.5%) | 428 (31.6%) | 572 (30.2%) |
| 3 | 204 (35.1%) | 121 (29.4%) | 221 (31.3%) | 177 (27.8%) | 107 (24.2%) | 186 (30.0%) | 124 (19.7%) | 100 (19.9%) | 149 (26.3%) | 505 (27.3%) | 328 (24.2%) | 556 (29.4%) |
table1(~ factor(ab) | my3*Tx, data=qtr, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=582) |
A_SDCT (N=411) |
BDCT (N=706) |
C_SDCT (N=636) |
A_SDCT (N=442) |
BDCT (N=621) |
C_SDCT (N=631) |
A_SDCT (N=502) |
BDCT (N=566) |
C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
|
| factor(ab) | ||||||||||||
| 0 | 345 (59.3%) | 162 (39.4%) | 0 (0%) | 371 (58.3%) | 271 (61.3%) | 0 (0%) | 333 (52.8%) | 326 (64.9%) | 0 (0%) | 1049 (56.7%) | 759 (56.0%) | 0 (0%) |
| 1 | 237 (40.7%) | 249 (60.6%) | 706 (100%) | 265 (41.7%) | 171 (38.7%) | 621 (100%) | 298 (47.2%) | 176 (35.1%) | 566 (100%) | 800 (43.3%) | 596 (44.0%) | 1893 (100%) |
table1(~ factor(imipc) + factor(cure) + factor(newimi) | my3*Tx, data=qtr, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=582) |
A_SDCT (N=411) |
BDCT (N=706) |
C_SDCT (N=636) |
A_SDCT (N=442) |
BDCT (N=621) |
C_SDCT (N=631) |
A_SDCT (N=502) |
BDCT (N=566) |
C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
|
| factor(imipc) | ||||||||||||
| 0 | 440 (75.6%) | 326 (79.3%) | 568 (80.5%) | 504 (79.2%) | 340 (76.9%) | 497 (80.0%) | 481 (76.2%) | 386 (76.9%) | 435 (76.9%) | 1425 (77.1%) | 1052 (77.6%) | 1500 (79.2%) |
| 1 | 142 (24.4%) | 85 (20.7%) | 138 (19.5%) | 132 (20.8%) | 102 (23.1%) | 124 (20.0%) | 150 (23.8%) | 116 (23.1%) | 131 (23.1%) | 424 (22.9%) | 303 (22.4%) | 393 (20.8%) |
| factor(cure) | ||||||||||||
| 0 | 13 (11.6%) | 8 (11.8%) | 22 (13.7%) | 20 (13.2%) | 13 (12.6%) | 13 (9.49%) | 16 (9.36%) | 15 (11.4%) | 22 (14.5%) | 49 (11.3%) | 36 (11.9%) | 57 (12.7%) |
| 1 | 99 (88.4%) | 60 (88.2%) | 139 (86.3%) | 131 (86.8%) | 90 (87.4%) | 124 (90.5%) | 155 (90.6%) | 117 (88.6%) | 130 (85.5%) | 385 (88.7%) | 267 (88.1%) | 393 (87.3%) |
| Missing | 470 (80.8%) | 343 (83.5%) | 545 (77.2%) | 485 (76.3%) | 339 (76.7%) | 484 (77.9%) | 460 (72.9%) | 370 (73.7%) | 414 (73.1%) | 1415 (76.5%) | 1052 (77.6%) | 1443 (76.2%) |
| factor(newimi) | ||||||||||||
| 0 | 416 (76.2%) | 295 (78.7%) | 530 (81.2%) | 472 (81.2%) | 318 (79.3%) | 463 (81.4%) | 435 (76.9%) | 351 (78.3%) | 393 (78.8%) | 1323 (78.1%) | 964 (78.8%) | 1386 (80.5%) |
| 1 | 130 (23.8%) | 80 (21.3%) | 123 (18.8%) | 109 (18.8%) | 83 (20.7%) | 106 (18.6%) | 131 (23.1%) | 97 (21.7%) | 106 (21.2%) | 370 (21.9%) | 260 (21.2%) | 335 (19.5%) |
| Missing | 36 (6.2%) | 36 (8.8%) | 53 (7.5%) | 55 (8.6%) | 41 (9.3%) | 52 (8.4%) | 65 (10.3%) | 54 (10.8%) | 67 (11.8%) | 156 (8.4%) | 131 (9.7%) | 172 (9.1%) |
table1(~ factor(CM) + factor(CMDP) + factor(Parity) | Tx, data=cow, na.is.category = F)| C_SDCT (N=536) |
A_SDCT (N=394) |
BDCT (N=555) |
Overall (N=1485) |
|
|---|---|---|---|---|
| factor(CM) | ||||
| 0 | 472 (88.1%) | 346 (87.8%) | 464 (83.6%) | 1282 (86.3%) |
| 1 | 64 (11.9%) | 48 (12.2%) | 91 (16.4%) | 203 (13.7%) |
| factor(CMDP) | ||||
| 0 | 534 (99.6%) | 391 (99.2%) | 554 (99.8%) | 1479 (99.6%) |
| 1 | 2 (0.373%) | 3 (0.761%) | 1 (0.180%) | 6 (0.404%) |
| factor(Parity) | ||||
| 1 | 241 (45.0%) | 167 (42.4%) | 218 (39.3%) | 626 (42.2%) |
| 2 | 141 (26.3%) | 123 (31.2%) | 170 (30.6%) | 434 (29.2%) |
| 3 | 154 (28.7%) | 104 (26.4%) | 167 (30.1%) | 425 (28.6%) |
table1(~ factor(CM) + factor(CMDP) | my3*Tx, data=cow, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=171) |
A_SDCT (N=121) |
BDCT (N=203) |
C_SDCT (N=182) |
A_SDCT (N=127) |
BDCT (N=186) |
C_SDCT (N=183) |
A_SDCT (N=146) |
BDCT (N=166) |
C_SDCT (N=536) |
A_SDCT (N=394) |
BDCT (N=555) |
|
| factor(CM) | ||||||||||||
| 0 | 154 (90.1%) | 102 (84.3%) | 166 (81.8%) | 157 (86.3%) | 113 (89.0%) | 161 (86.6%) | 161 (88.0%) | 131 (89.7%) | 137 (82.5%) | 472 (88.1%) | 346 (87.8%) | 464 (83.6%) |
| 1 | 17 (9.94%) | 19 (15.7%) | 37 (18.2%) | 25 (13.7%) | 14 (11.0%) | 25 (13.4%) | 22 (12.0%) | 15 (10.3%) | 29 (17.5%) | 64 (11.9%) | 48 (12.2%) | 91 (16.4%) |
| factor(CMDP) | ||||||||||||
| 0 | 170 (99.4%) | 121 (100%) | 203 (100%) | 182 (100%) | 127 (100%) | 186 (100%) | 182 (99.5%) | 143 (97.9%) | 165 (99.4%) | 534 (99.6%) | 391 (99.2%) | 554 (99.8%) |
| 1 | 1 (0.585%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 1 (0.546%) | 3 (2.05%) | 1 (0.602%) | 2 (0.373%) | 3 (0.761%) | 1 (0.180%) |
cow %>% tabyl(source,my3) %>% adorn_totals("row") %>%
adorn_percentages("row") %>%
adorn_pct_formatting(digits = 1) %>%
adorn_ns %>%
adorn_title## my3
## source Low production Mid production High production
## CA 48.9% (134) 35.0% (96) 16.1% (44)
## US 29.8% (361) 32.9% (399) 37.2% (451)
## Total 33.3% (495) 33.3% (495) 33.3% (495)cow %>% tabyl(FARMID,my3) %>% adorn_totals("row") %>%
adorn_percentages("row") %>%
adorn_pct_formatting(digits = 1) %>%
adorn_ns %>%
adorn_title## my3
## FARMID Low production Mid production High production
## 1 56.6% (43) 31.6% (24) 11.8% (9)
## 2 31.7% (26) 28.0% (23) 40.2% (33)
## 3 20.5% (54) 41.4% (109) 38.0% (100)
## 4 18.3% (71) 29.1% (113) 52.6% (204)
## 5 60.7% (147) 30.2% (73) 9.1% (22)
## 6 6.4% (3) 38.3% (18) 55.3% (26)
## 7 15.0% (17) 34.5% (39) 50.4% (57)
## 8 72.2% (13) 16.7% (3) 11.1% (2)
## 9 76.0% (19) 24.0% (6) 0.0% (0)
## 10 17.4% (4) 60.9% (14) 21.7% (5)
## 11 50.0% (11) 31.8% (7) 18.2% (4)
## 12 78.9% (15) 15.8% (3) 5.3% (1)
## 13 69.0% (20) 27.6% (8) 3.4% (1)
## 14 36.8% (28) 34.2% (26) 28.9% (22)
## 15 31.2% (5) 43.8% (7) 25.0% (4)
## 16 41.3% (19) 47.8% (22) 10.9% (5)
## Total 33.3% (495) 33.3% (495) 33.3% (495)1st lactation cows are over-represented in the Algorithm cows, probably due to
Extract crude frequencies
cm.freq.tx <- cow %>% group_by(Tx) %>% summarise(
n.cm.0 = sum(CM==0),
n.cm.1 = sum(CM==1),
p.cm.1 = n.cm.1/(n.cm.1+n.cm.0),
cm = paste0(n.cm.1, "/" ,n.cm.0 + n.cm.1, " (",100*round(p.cm.1,4),"%)")
)
cm.freq.tx$my3 <- "All production levels"
cm.freq.my3 <- cow %>% group_by(my3,Tx) %>% summarise(
n.cm.0 = sum(CM==0),
n.cm.1 = sum(CM==1),
p.cm.1 = n.cm.1/(n.cm.1+n.cm.0),
cm = paste0(n.cm.1, "/" ,n.cm.0 + n.cm.1, " (",100*round(p.cm.1,4),"%)")
)
cm.freq <- rbind(cm.freq.tx,cm.freq.my3) %>% select(my3,Tx,cm)
cm.freq$risk <- cm.freq$cm
cm.freq$outcome <- "Clinical mastitis"
cm.freq$cm <- NULL
#Cull
Cull2.freq.tx <- cow %>% group_by(Tx) %>% summarise(
n.Cull2.0 = sum(Cull2==0),
n.Cull2.1 = sum(Cull2==1),
p.Cull2.1 = n.Cull2.1/(n.Cull2.1+n.Cull2.0),
Cull2 = paste0(n.Cull2.1, "/" ,n.Cull2.0 + n.Cull2.1, " (",100*round(p.Cull2.1,4),"%)")
)
Cull2.freq.tx$my3 <- "All production levels"
Cull2.freq.my3 <- cow %>% group_by(my3,Tx) %>% summarise(
n.Cull2.0 = sum(Cull2==0),
n.Cull2.1 = sum(Cull2==1),
p.Cull2.1 = n.Cull2.1/(n.Cull2.1+n.Cull2.0),
Cull2 = paste0(n.Cull2.1, "/" ,n.Cull2.0 + n.Cull2.1, " (",100*round(p.Cull2.1,4),"%)")
)
Cull2.freq <- rbind(Cull2.freq.tx,Cull2.freq.my3) %>% select(my3,Tx,Cull2)
Cull2.freq$risk <- Cull2.freq$Cull2
Cull2.freq$outcome <- "Cull"
Cull2.freq$Cull2 <- NULL
imi.freq.tx <- qtr %>% group_by(Tx) %>% summarise(
n.imipc.0 = sum(imipc==0),
n.imipc.1 = sum(imipc==1),
p.imipc.1 = n.imipc.1/(n.imipc.1+n.imipc.0),
imipc = paste0(n.imipc.1, "/" ,n.imipc.0 + n.imipc.1, " (",100*round(p.imipc.1,4),"%)"),
n.newimi.0 = sum(newimi==0,na.rm=T),
n.newimi.1 = sum(newimi==1,na.rm=T),
p.newimi.1 = n.newimi.1/(n.newimi.1+n.newimi.0),
newimi = paste0(n.newimi.1, "/" ,n.newimi.0 + n.newimi.1, " (",100*round(p.newimi.1,4),"%)"),
n.cure.0 = sum(cure==0,na.rm=T),
n.cure.1 = sum(cure==1,na.rm=T),
p.cure.1 = n.cure.1/(n.cure.1+n.cure.0),
imicure = paste0(n.cure.1, "/" ,n.cure.0 + n.cure.1, " (",100*round(p.cure.1,4),"%)"),
)
imi.freq.tx$my3 <- "All production levels"
imi.freq.my3 <- qtr %>% group_by(my3,Tx) %>% summarise(
n.imipc.0 = sum(imipc==0),
n.imipc.1 = sum(imipc==1),
p.imipc.1 = n.imipc.1/(n.imipc.1+n.imipc.0),
imipc = paste0(n.imipc.1, "/" ,n.imipc.0 + n.imipc.1, " (",100*round(p.imipc.1,4),"%)"),
n.newimi.0 = sum(newimi==0,na.rm=T),
n.newimi.1 = sum(newimi==1,na.rm=T),
p.newimi.1 = n.newimi.1/(n.newimi.1+n.newimi.0),
newimi = paste0(n.newimi.1, "/" ,n.newimi.0 + n.newimi.1, " (",100*round(p.newimi.1,4),"%)"),
n.cure.0 = sum(cure==0,na.rm=T),
n.cure.1 = sum(cure==1,na.rm=T),
p.cure.1 = n.cure.1/(n.cure.1+n.cure.0),
imicure = paste0(n.cure.1, "/" ,n.cure.0 + n.cure.1, " (",100*round(p.cure.1,4),"%)"),
)
colnames(imi.freq.tx)## [1] "Tx" "n.imipc.0" "n.imipc.1" "p.imipc.1" "imipc"
## [6] "n.newimi.0" "n.newimi.1" "p.newimi.1" "newimi" "n.cure.0"
## [11] "n.cure.1" "p.cure.1" "imicure" "my3"colnames(imi.freq.my3)## [1] "my3" "Tx" "n.imipc.0" "n.imipc.1" "p.imipc.1"
## [6] "imipc" "n.newimi.0" "n.newimi.1" "p.newimi.1" "newimi"
## [11] "n.cure.0" "n.cure.1" "p.cure.1" "imicure"imi.freq <- rbind(imi.freq.tx,imi.freq.my3) %>% select(my3,Tx,imipc,newimi,imicure) %>% pivot_longer(
cols = imipc:imicure,
names_to = "outcome",
values_to = "risk"
)
freq.tab <- rbind(imi.freq,cm.freq,Cull2.freq) %>% arrange(outcome,my3)
write.csv(freq.tab,"freq.table.csv")Analysis
Post calving clinical mastitis
library(survival)
library(survminer)
KM <- survfit(Surv(CMTAR, CM) ~ 1, data = cow)
cmtable <- summary(KM, times = c(30,60,90,120*(1:10)))
data.frame(
days = cmtable[["time"]],
incidence = (1 - cmtable[["surv"]])*100
)## days incidence
## 1 30 3.759954
## 2 60 7.614181
## 3 90 10.809019
## 4 120 14.249354ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
Comparison of US and Canadian clinical mastitis incidence
KM <- survfit(Surv(CMTAR, CM) ~ source, data = cow)
ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow)
#summary(KM)
ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
Clinical mastitis - all production groups combined
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow)
km.all <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Low production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="Low production"))
km.low <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "Low production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Mid production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="Mid production"))
km.mid <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "Mid production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)High production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="High production"))
km.high <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "High production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
legend.cm <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "bottom",
title = "High production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Export images
splots <- list(km.low,
km.mid,
km.high)
tiff("KM.tiff", units="in", width=13*.7, height=4*.7, res=300)
arrange_ggsurvplots(splots,ncol=3,nrow=1)
dev.off()## quartz_off_screen
## 2arrange_ggsurvplots(splots,ncol=3,nrow=1)
tiff("legend.tiff", units="in", width=6, height=3, res=300)
legend.cm
dev.off()## quartz_off_screen
## 2Cox regression
All production groups pooled together
library(broom)
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
SR <- coxph(Surv(CMTAR, CM) ~ Tx + my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(SR)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(CMTAR, CM)
## Df Chisq Pr(>Chisq)
## Tx 2 4.3132 0.11572
## my3 2 0.3167 0.85355
## factor(Parity) 2 7.2864 0.02617 *
## DOSCC 1 1.1321 0.28732
## factor(source) 1 1.1883 0.27567
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(SR)## Call:
## coxph(formula = Surv(CMTAR, CM) ~ Tx + my3 + factor(Parity) +
## DOSCC + factor(source), data = cow, cluster = FARMID)
##
## n= 1484, number of events= 203
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z Pr(>|z|)
## TxC_SDCT -0.31922 0.72672 0.16346 0.17060 -1.871 0.06132 .
## TxA_SDCT -0.22858 0.79567 0.18654 0.14708 -1.554 0.12016
## my3Mid production -0.02463 0.97567 0.17890 0.23770 -0.104 0.91746
## my3High production 0.04946 1.05070 0.18300 0.30945 0.160 0.87302
## factor(Parity)2 0.31451 1.36959 0.17436 0.11654 2.699 0.00696 **
## factor(Parity)3 0.28821 1.33403 0.17961 0.23617 1.220 0.22235
## DOSCC 0.10164 1.10698 0.06410 0.09552 1.064 0.28732
## factor(source)US -0.25687 0.77347 0.17972 0.23564 -1.090 0.27567
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 0.7267 1.3760 0.5202 1.015
## TxA_SDCT 0.7957 1.2568 0.5964 1.062
## my3Mid production 0.9757 1.0249 0.6123 1.555
## my3High production 1.0507 0.9517 0.5729 1.927
## factor(Parity)2 1.3696 0.7301 1.0899 1.721
## factor(Parity)3 1.3340 0.7496 0.8397 2.119
## DOSCC 1.1070 0.9034 0.9180 1.335
## factor(source)US 0.7735 1.2929 0.4874 1.227
##
## Concordance= 0.581 (se = 0.031 )
## Likelihood ratio test= 16.62 on 8 df, p=0.03
## Wald test = 52.13 on 8 df, p=2e-08
## Score (logrank) test = 16.81 on 8 df, p=0.03, Robust = 10.38 p=0.2
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).cm.main <- tidy(SR,conf.int = T,exp=T)[1:2,] %>% select(term,estimate,conf.low,conf.high)
cm.main <- cm.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
cm.main$production <- "All production levels"
cm.main$contrast <- "none"
cm.main$outcome <- "Clinical mastitis"Model diagnostics
SR <- cox.zph(SR)
SR## chisq df p
## Tx 3.283 2 0.19
## my3 0.799 2 0.67
## factor(Parity) 2.112 2 0.35
## DOSCC 0.959 1 0.33
## factor(source) 1.524 1 0.22
## GLOBAL 9.887 8 0.27ggcoxzph(SR)
Interaction model
CPH <- coxph(Surv(CMTAR, CM) ~ Tx*my3 + Parity + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(CMTAR, CM)
## Df Chisq Pr(>Chisq)
## Tx 2 9.0584 0.01079 *
## my3 2 0.0731 0.96412
## Parity 2 6.3562 0.04166 *
## DOSCC 1 1.2109 0.27116
## factor(source) 1 1.5412 0.21444
## Tx:my3 4 11.8755 0.01830 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(CPH)## Call:
## coxph(formula = Surv(CMTAR, CM) ~ Tx + my3 + Parity + DOSCC +
## factor(source) + Tx:my3, data = cow, cluster = FARMID)
##
## n= 1484, number of events= 203
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z
## TxC_SDCT -0.65608 0.51888 0.29329 0.43924 -1.494
## TxA_SDCT -0.01085 0.98921 0.29136 0.26769 -0.041
## my3Mid production -0.19858 0.81989 0.26454 0.32092 -0.619
## my3High production 0.08911 1.09320 0.25892 0.40401 0.221
## Parity2 0.30233 1.35301 0.17451 0.12005 2.518
## Parity3 0.27770 1.32009 0.17976 0.23911 1.161
## DOSCC 0.10641 1.11227 0.06430 0.09670 1.100
## factor(source)US -0.29858 0.74187 0.18109 0.24051 -1.241
## TxC_SDCT:my3Mid production 0.67550 1.96501 0.40779 0.44977 1.502
## TxA_SDCT:my3Mid production -0.13496 0.87375 0.43753 0.38804 -0.348
## TxC_SDCT:my3High production 0.32126 1.37886 0.40764 0.51713 0.621
## TxA_SDCT:my3High production -0.47544 0.62161 0.42788 0.39897 -1.192
## Pr(>|z|)
## TxC_SDCT 0.1353
## TxA_SDCT 0.9677
## my3Mid production 0.5361
## my3High production 0.8254
## Parity2 0.0118 *
## Parity3 0.2455
## DOSCC 0.2712
## factor(source)US 0.2144
## TxC_SDCT:my3Mid production 0.1331
## TxA_SDCT:my3Mid production 0.7280
## TxC_SDCT:my3High production 0.5344
## TxA_SDCT:my3High production 0.2334
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 0.5189 1.9272 0.2194 1.227
## TxA_SDCT 0.9892 1.0109 0.5854 1.672
## my3Mid production 0.8199 1.2197 0.4371 1.538
## my3High production 1.0932 0.9147 0.4952 2.413
## Parity2 1.3530 0.7391 1.0693 1.712
## Parity3 1.3201 0.7575 0.8262 2.109
## DOSCC 1.1123 0.8991 0.9202 1.344
## factor(source)US 0.7419 1.3479 0.4630 1.189
## TxC_SDCT:my3Mid production 1.9650 0.5089 0.8138 4.745
## TxA_SDCT:my3Mid production 0.8737 1.1445 0.4084 1.869
## TxC_SDCT:my3High production 1.3789 0.7252 0.5004 3.799
## TxA_SDCT:my3High production 0.6216 1.6087 0.2844 1.359
##
## Concordance= 0.591 (se = 0.034 )
## Likelihood ratio test= 21.68 on 12 df, p=0.04
## Wald test = 83.14 on 12 df, p=1e-12
## Score (logrank) test = 21.85 on 12 df, p=0.04, Robust = 14.02 p=0.3
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).emmeans(CPH,pairwise ~ Tx | my3,adjust = "none",type = "response") %>% confint## $emmeans
## my3 = Low production:
## Tx response SE df asymp.LCL asymp.UCL
## BDCT 1.675 0.658 Inf 0.775 3.62
## C_SDCT 0.869 0.571 Inf 0.240 3.15
## A_SDCT 1.657 0.946 Inf 0.541 5.07
##
## my3 = Mid production:
## Tx response SE df asymp.LCL asymp.UCL
## BDCT 1.373 0.787 Inf 0.446 4.22
## C_SDCT 1.400 0.718 Inf 0.512 3.82
## A_SDCT 1.187 0.467 Inf 0.549 2.57
##
## my3 = High production:
## Tx response SE df asymp.LCL asymp.UCL
## BDCT 1.831 1.009 Inf 0.621 5.39
## C_SDCT 1.310 0.913 Inf 0.334 5.13
## A_SDCT 1.126 0.491 Inf 0.479 2.65
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the log scale
##
## $contrasts
## my3 = Low production:
## contrast ratio SE df asymp.LCL asymp.UCL
## BDCT / C_SDCT 1.927 0.847 Inf 0.815 4.56
## BDCT / A_SDCT 1.011 0.271 Inf 0.598 1.71
## C_SDCT / A_SDCT 0.525 0.185 Inf 0.263 1.05
##
## my3 = Mid production:
## contrast ratio SE df asymp.LCL asymp.UCL
## BDCT / C_SDCT 0.981 0.193 Inf 0.667 1.44
## BDCT / A_SDCT 1.157 0.282 Inf 0.718 1.86
## C_SDCT / A_SDCT 1.180 0.277 Inf 0.744 1.87
##
## my3 = High production:
## contrast ratio SE df asymp.LCL asymp.UCL
## BDCT / C_SDCT 1.398 0.430 Inf 0.764 2.56
## BDCT / A_SDCT 1.626 0.428 Inf 0.971 2.72
## C_SDCT / A_SDCT 1.164 0.486 Inf 0.513 2.64
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the log scaleModel diagnostics
SR <- cox.zph(CPH)
SR## chisq df p
## Tx 3.243 2 0.20
## my3 0.790 2 0.67
## Parity 2.088 2 0.35
## DOSCC 0.981 1 0.32
## factor(source) 1.546 1 0.21
## Tx:my3 0.103 4 1.00
## GLOBAL 17.871 12 0.12ggcoxzph(SR)
cow$Tx <- fct_relevel(cow$Tx,"C_SDCT","A_SDCT","BDCT")
CPH <- coxph(Surv(CMTAR, CM) ~ Tx*my3 + strata(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
out <- emmeans(CPH,pairwise ~ Tx | my3,adjust = "none",type = "response") %>% confint
CPH2 <- coxph(Surv(CMTAR, CM) ~ cult*my3 + alg*my3 + strata(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
int.p <- car::Anova(CPH2) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "Clinical mastitis"
int.p.cm <- int.p
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["ratio"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT / A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Clinical mastitis"
tab.cm <- tab
tab.cm## production contrast estimate outcome
## 1 Low production C_SDCT / BDCT 0.52 (0.22 to 1.23) Clinical mastitis
## 2 Low production A_SDCT / BDCT 0.99 (0.59 to 1.66) Clinical mastitis
## 3 Mid production C_SDCT / BDCT 1.02 (0.69 to 1.5) Clinical mastitis
## 4 Mid production A_SDCT / BDCT 0.87 (0.54 to 1.39) Clinical mastitis
## 5 High production C_SDCT / BDCT 0.72 (0.39 to 1.31) Clinical mastitis
## 6 High production A_SDCT / BDCT 0.62 (0.37 to 1.03) Clinical mastitisCulling 1-120 DIM
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
KM <- survfit(Surv(Cull2TAR, Cull2) ~ 1, data = cow)
cmtable <- summary(KM, times = c(30,60,90,120*(1:10)))
data.frame(
days = cmtable[["time"]],
incidence = (1 - cmtable[["surv"]])*100
)## days incidence
## 1 30 3.771044
## 2 60 6.127946
## 3 90 7.946128
## 4 120 9.966330ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
summary(KM)## Call: survfit(formula = Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
##
## Tx=BDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 3 555 1 0.998 0.00180 0.995 1.000
## 4 554 1 0.996 0.00254 0.991 1.000
## 5 553 1 0.995 0.00311 0.989 1.000
## 6 552 2 0.991 0.00401 0.983 0.999
## 7 550 2 0.987 0.00474 0.978 0.997
## 9 548 1 0.986 0.00506 0.976 0.996
## 11 547 2 0.982 0.00565 0.971 0.993
## 13 545 2 0.978 0.00617 0.966 0.991
## 15 543 1 0.977 0.00642 0.964 0.989
## 19 542 2 0.973 0.00688 0.960 0.987
## 21 540 2 0.969 0.00731 0.955 0.984
## 22 538 2 0.966 0.00772 0.951 0.981
## 23 536 2 0.962 0.00810 0.946 0.978
## 27 534 2 0.959 0.00846 0.942 0.975
## 28 532 2 0.955 0.00880 0.938 0.972
## 29 530 1 0.953 0.00897 0.936 0.971
## 30 529 1 0.951 0.00913 0.934 0.969
## 36 528 2 0.948 0.00945 0.929 0.966
## 38 526 1 0.946 0.00960 0.927 0.965
## 39 525 1 0.944 0.00975 0.925 0.963
## 41 524 2 0.941 0.01004 0.921 0.960
## 49 522 2 0.937 0.01032 0.917 0.957
## 54 520 1 0.935 0.01045 0.915 0.956
## 61 519 1 0.933 0.01059 0.913 0.954
## 64 518 1 0.932 0.01072 0.911 0.953
## 65 517 1 0.930 0.01085 0.909 0.951
## 72 516 1 0.928 0.01098 0.907 0.950
## 75 515 1 0.926 0.01110 0.905 0.948
## 78 514 1 0.924 0.01123 0.903 0.947
## 79 513 1 0.923 0.01135 0.901 0.945
## 83 512 1 0.921 0.01147 0.899 0.943
## 93 511 2 0.917 0.01170 0.894 0.940
## 99 509 2 0.914 0.01193 0.890 0.937
## 101 507 1 0.912 0.01204 0.888 0.936
## 112 506 1 0.910 0.01215 0.886 0.934
## 115 505 1 0.908 0.01226 0.884 0.932
## 116 504 1 0.906 0.01237 0.882 0.931
## 117 503 1 0.905 0.01248 0.880 0.929
## 118 502 1 0.903 0.01258 0.878 0.928
##
## Tx=C_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 1 536 1 0.998 0.00186 0.994 1.000
## 2 535 2 0.994 0.00322 0.988 1.000
## 3 533 2 0.991 0.00415 0.983 0.999
## 4 531 2 0.987 0.00490 0.977 0.997
## 5 529 1 0.985 0.00524 0.975 0.995
## 7 528 1 0.983 0.00555 0.972 0.994
## 9 527 1 0.981 0.00584 0.970 0.993
## 11 526 1 0.979 0.00612 0.968 0.992
## 13 525 1 0.978 0.00639 0.965 0.990
## 21 524 1 0.976 0.00664 0.963 0.989
## 23 523 1 0.974 0.00689 0.960 0.987
## 24 522 1 0.972 0.00712 0.958 0.986
## 33 521 1 0.970 0.00735 0.956 0.985
## 34 520 1 0.968 0.00757 0.954 0.983
## 36 519 1 0.966 0.00778 0.951 0.982
## 37 518 2 0.963 0.00819 0.947 0.979
## 38 516 1 0.961 0.00838 0.945 0.977
## 42 515 1 0.959 0.00857 0.942 0.976
## 43 514 1 0.957 0.00875 0.940 0.974
## 44 513 2 0.953 0.00911 0.936 0.971
## 45 511 1 0.951 0.00928 0.933 0.970
## 49 510 1 0.950 0.00945 0.931 0.968
## 57 509 1 0.948 0.00961 0.929 0.967
## 58 508 1 0.946 0.00977 0.927 0.965
## 63 507 1 0.944 0.00993 0.925 0.964
## 69 506 1 0.942 0.01008 0.923 0.962
## 70 505 1 0.940 0.01023 0.920 0.961
## 74 504 1 0.938 0.01038 0.918 0.959
## 75 503 1 0.937 0.01053 0.916 0.957
## 76 502 1 0.935 0.01067 0.914 0.956
## 79 501 1 0.933 0.01081 0.912 0.954
## 84 500 1 0.931 0.01095 0.910 0.953
## 89 499 1 0.929 0.01109 0.908 0.951
## 91 498 2 0.925 0.01135 0.903 0.948
## 96 496 2 0.922 0.01161 0.899 0.945
## 99 494 1 0.920 0.01173 0.897 0.943
## 111 493 2 0.916 0.01198 0.893 0.940
## 114 491 1 0.914 0.01210 0.891 0.938
## 115 490 1 0.912 0.01222 0.889 0.937
## 116 489 2 0.909 0.01245 0.885 0.933
## 117 487 2 0.905 0.01267 0.880 0.930
## 119 485 1 0.903 0.01278 0.878 0.928
##
## Tx=A_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 1 394 2 0.995 0.00358 0.988 1.000
## 2 392 1 0.992 0.00438 0.984 1.000
## 3 391 1 0.990 0.00505 0.980 1.000
## 4 390 1 0.987 0.00564 0.976 0.998
## 6 389 1 0.985 0.00617 0.973 0.997
## 9 388 1 0.982 0.00666 0.969 0.995
## 13 387 3 0.975 0.00792 0.959 0.990
## 18 384 1 0.972 0.00830 0.956 0.988
## 19 383 1 0.970 0.00866 0.953 0.987
## 21 382 1 0.967 0.00900 0.950 0.985
## 24 381 1 0.964 0.00933 0.946 0.983
## 33 380 1 0.962 0.00964 0.943 0.981
## 36 379 1 0.959 0.00994 0.940 0.979
## 41 378 1 0.957 0.01024 0.937 0.977
## 42 377 1 0.954 0.01052 0.934 0.975
## 44 376 1 0.952 0.01079 0.931 0.973
## 45 375 1 0.949 0.01106 0.928 0.971
## 51 374 1 0.947 0.01132 0.925 0.969
## 52 373 1 0.944 0.01157 0.922 0.967
## 54 372 1 0.942 0.01181 0.919 0.965
## 55 371 1 0.939 0.01205 0.916 0.963
## 58 370 1 0.937 0.01228 0.913 0.961
## 60 369 1 0.934 0.01251 0.910 0.959
## 62 368 1 0.931 0.01273 0.907 0.957
## 74 367 1 0.929 0.01294 0.904 0.955
## 77 366 2 0.924 0.01336 0.898 0.950
## 81 364 1 0.921 0.01356 0.895 0.948
## 85 363 1 0.919 0.01376 0.892 0.946
## 86 362 1 0.916 0.01396 0.889 0.944
## 87 361 1 0.914 0.01415 0.886 0.942
## 88 360 1 0.911 0.01433 0.884 0.940
## 90 359 1 0.909 0.01452 0.881 0.938
## 97 358 1 0.906 0.01470 0.878 0.935
## 99 357 1 0.904 0.01487 0.875 0.933
## 102 356 1 0.901 0.01505 0.872 0.931
## 105 355 1 0.898 0.01522 0.869 0.929
## 111 354 1 0.896 0.01538 0.866 0.927
## 116 353 1 0.893 0.01555 0.863 0.924ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
Culling - all production groups combined
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
km.all <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Low production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="Low production"))
km.low <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "Low production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Mid production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="Mid production"))
km.mid <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "Mid production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)High production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="High production"))
km.high <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),legend.labs = c("Blanket DCT","Culture SDCT","Algorithm SDCT"),
legend = "none",
title = "High production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Export images
splots <- list(km.low,
km.mid,
km.high)
tiff("KM2.tiff", units="in", width=13*.7, height=4*.7, res=300)
arrange_ggsurvplots(splots,ncol=3,nrow=1)
dev.off()## quartz_off_screen
## 2arrange_ggsurvplots(splots,ncol=3,nrow=1)
Cox regression
All production groups pooled together
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
SR <- coxph(Surv(Cull2TAR, Cull2) ~ Tx + my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(SR)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 0.2417 0.886185
## my3 2 3.6772 0.159039
## factor(Parity) 2 28.1800 7.6e-07 ***
## DOSCC 1 10.0458 0.001527 **
## factor(source) 1 5.3009 0.021314 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(SR)## Call:
## coxph(formula = Surv(Cull2TAR, Cull2) ~ Tx + my3 + factor(Parity) +
## DOSCC + factor(source), data = cow, cluster = FARMID)
##
## n= 1484, number of events= 148
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z Pr(>|z|)
## TxC_SDCT 0.01656 1.01669 0.19486 0.17473 0.095 0.92452
## TxA_SDCT 0.05923 1.06101 0.21274 0.21555 0.275 0.78350
## my3Mid production -0.29969 0.74105 0.20648 0.23807 -1.259 0.20810
## my3High production -0.36079 0.69712 0.21546 0.18815 -1.918 0.05517 .
## factor(Parity)2 0.25084 1.28510 0.23463 0.18764 1.337 0.18129
## factor(Parity)3 0.96662 2.62904 0.20835 0.20476 4.721 2.35e-06 ***
## DOSCC 0.12146 1.12915 0.07417 0.03832 3.170 0.00153 **
## factor(source)US 0.35858 1.43129 0.24290 0.15574 2.302 0.02131 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 1.0167 0.9836 0.7219 1.432
## TxA_SDCT 1.0610 0.9425 0.6954 1.619
## my3Mid production 0.7410 1.3494 0.4647 1.182
## my3High production 0.6971 1.4345 0.4821 1.008
## factor(Parity)2 1.2851 0.7781 0.8896 1.856
## factor(Parity)3 2.6290 0.3804 1.7600 3.927
## DOSCC 1.1291 0.8856 1.0474 1.217
## factor(source)US 1.4313 0.6987 1.0548 1.942
##
## Concordance= 0.648 (se = 0.022 )
## Likelihood ratio test= 45.04 on 8 df, p=4e-07
## Wald test = 454.2 on 8 df, p=<2e-16
## Score (logrank) test = 48.55 on 8 df, p=8e-08, Robust = 9.11 p=0.3
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).cull.main <- tidy(SR,conf.int = T,exp=T)[1:2,] %>% select(term,estimate,conf.low,conf.high)
cull.main <- cull.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
cull.main$production <- "All production levels"
cull.main$contrast <- "none"
cull.main$outcome <- "Culling"Model diagnostics
SR <- cox.zph(SR)
SR## chisq df p
## Tx 1.403 2 0.50
## my3 2.390 2 0.30
## factor(Parity) 1.159 2 0.56
## DOSCC 0.772 1 0.38
## factor(source) 1.175 1 0.28
## GLOBAL 5.768 8 0.67ggcoxzph(SR)
Interaction model
CPH <- coxph(Surv(Cull2TAR, Cull2) ~ Tx*my3 + Parity + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 3.0951 0.212771
## my3 2 83.0898 < 2.2e-16 ***
## Parity 2 26.0526 2.202e-06 ***
## DOSCC 1 10.8066 0.001011 **
## factor(source) 1 4.9243 0.026481 *
## Tx:my3 4 3.2706 0.513606
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(CPH)## Call:
## coxph(formula = Surv(Cull2TAR, Cull2) ~ Tx + my3 + Parity + DOSCC +
## factor(source) + Tx:my3, data = cow, cluster = FARMID)
##
## n= 1484, number of events= 148
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z
## TxC_SDCT 0.02307 1.02334 0.29911 0.36641 0.063
## TxA_SDCT 0.24488 1.27747 0.31050 0.22565 1.085
## my3Mid production -0.13088 0.87732 0.32430 0.45901 -0.285
## my3High production -0.33376 0.71622 0.35523 0.39258 -0.850
## Parity2 0.24447 1.27694 0.23497 0.18706 1.307
## Parity3 0.96071 2.61354 0.20869 0.21027 4.569
## DOSCC 0.12487 1.13300 0.07423 0.03799 3.287
## factor(source)US 0.34701 1.41483 0.24425 0.15638 2.219
## TxC_SDCT:my3Mid production -0.10820 0.89745 0.45934 0.62570 -0.173
## TxA_SDCT:my3Mid production -0.46963 0.62524 0.50106 0.42943 -1.094
## TxC_SDCT:my3High production 0.08828 1.09229 0.48296 0.63324 0.139
## TxA_SDCT:my3High production -0.19083 0.82627 0.50445 0.63638 -0.300
## Pr(>|z|)
## TxC_SDCT 0.94980
## TxA_SDCT 0.27783
## my3Mid production 0.77554
## my3High production 0.39522
## Parity2 0.19126
## Parity3 4.9e-06 ***
## DOSCC 0.00101 **
## factor(source)US 0.02648 *
## TxC_SDCT:my3Mid production 0.86271
## TxA_SDCT:my3Mid production 0.27413
## TxC_SDCT:my3High production 0.88913
## TxA_SDCT:my3High production 0.76428
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 1.0233 0.9772 0.4990 2.098
## TxA_SDCT 1.2775 0.7828 0.8209 1.988
## my3Mid production 0.8773 1.1398 0.3568 2.157
## my3High production 0.7162 1.3962 0.3318 1.546
## Parity2 1.2769 0.7831 0.8850 1.842
## Parity3 2.6135 0.3826 1.7308 3.946
## DOSCC 1.1330 0.8826 1.0517 1.221
## factor(source)US 1.4148 0.7068 1.0414 1.922
## TxC_SDCT:my3Mid production 0.8974 1.1143 0.2633 3.059
## TxA_SDCT:my3Mid production 0.6252 1.5994 0.2695 1.451
## TxC_SDCT:my3High production 1.0923 0.9155 0.3157 3.779
## TxA_SDCT:my3High production 0.8263 1.2103 0.2374 2.876
##
## Concordance= 0.647 (se = 0.022 )
## Likelihood ratio test= 46.09 on 12 df, p=7e-06
## Wald test = 787.2 on 12 df, p=<2e-16
## Score (logrank) test = 50.09 on 12 df, p=1e-06, Robust = 13.19 p=0.4
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).Model diagnostics
SR <- cox.zph(CPH)
SR## chisq df p
## Tx 1.414 2 0.49
## my3 2.378 2 0.30
## Parity 1.155 2 0.56
## DOSCC 0.754 1 0.39
## factor(source) 1.190 1 0.28
## Tx:my3 4.889 4 0.30
## GLOBAL 8.684 12 0.73ggcoxzph(SR)
cow$Tx <- fct_relevel(cow$Tx,"C_SDCT","A_SDCT","BDCT")
CPH <- coxph(Surv(Cull2TAR, Cull2) ~ Tx*my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 3.0951 0.212771
## my3 2 83.0898 < 2.2e-16 ***
## factor(Parity) 2 26.0526 2.202e-06 ***
## DOSCC 1 10.8066 0.001011 **
## factor(source) 1 4.9243 0.026481 *
## Tx:my3 4 3.2706 0.513606
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1out <- emmeans(CPH,pairwise ~ Tx | my3,adjust = "none",type = "response") %>% confint
CPH2 <- coxph(Surv(Cull2TAR, Cull2) ~ cult*my3 + alg*my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
int.p <- car::Anova(CPH2) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "Culling"
int.p.cull <- int.p
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["ratio"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT / A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Culling"
tab.cull <- tab
tab.cull## production contrast estimate outcome
## 1 Low production C_SDCT / BDCT 1.02 (0.5 to 2.1) Culling
## 2 Low production A_SDCT / BDCT 1.28 (0.82 to 1.99) Culling
## 3 Mid production C_SDCT / BDCT 0.92 (0.49 to 1.73) Culling
## 4 Mid production A_SDCT / BDCT 0.8 (0.44 to 1.46) Culling
## 5 High production C_SDCT / BDCT 1.12 (0.45 to 2.79) Culling
## 6 High production A_SDCT / BDCT 1.06 (0.29 to 3.79) CullingIMI post calving
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(imipc ~ Tx + my3 + Parity + factor(source) + DOSCC, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
imipc.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
imipc.main <- imipc.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
imipc.main$production <- "All production levels"
imipc.main$contrast <- "none"
imipc.main$outcome <- "IMI at calving"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(imipc ~ Tx*my3 + Parity + factor(source) + DOSCC, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
glm <- glm(imipc ~ alg*my3 + cult*my3 + Parity + factor(source) + DOSCC, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"))
int.p <- car::Anova(glm) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "imipc"
int.p.imipc <- int.p
summary(gee)##
## Call:
## geeglm(formula = imipc ~ Tx * my3 + Parity + factor(source) +
## DOSCC, family = binomial(link = "identity"), data = qtr %>%
## filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.058980 0.059952 0.968 0.32522
## TxA_SDCT -0.052961 0.026035 4.138 0.04193 *
## TxBDCT -0.049438 0.037557 1.733 0.18807
## my3Mid production -0.019865 0.026733 0.552 0.45742
## my3High production 0.012972 0.030122 0.185 0.66672
## Parity2 0.008496 0.022265 0.146 0.70279
## Parity3 0.018443 0.017062 1.168 0.27973
## factor(source)US 0.056618 0.044750 1.601 0.20579
## DOSCC 0.027901 0.009242 9.113 0.00254 **
## TxA_SDCT:my3Mid production 0.062130 0.040739 2.326 0.12724
## TxBDCT:my3Mid production 0.045203 0.036210 1.558 0.21190
## TxA_SDCT:my3High production 0.034456 0.041533 0.688 0.40677
## TxBDCT:my3High production 0.036414 0.033993 1.147 0.28408
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.1438
## Number of clusters: 16 Maximum cluster size: 1270emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.220 0.0283 Inf 0.1643 0.275
## A_SDCT Low production 0.167 0.0306 Inf 0.1068 0.227
## BDCT Low production 0.170 0.0365 Inf 0.0988 0.242
## C_SDCT Mid production 0.200 0.0255 Inf 0.1499 0.250
## A_SDCT Mid production 0.209 0.0249 Inf 0.1602 0.258
## BDCT Mid production 0.196 0.0327 Inf 0.1315 0.260
## C_SDCT High production 0.233 0.0212 Inf 0.1911 0.274
## A_SDCT High production 0.214 0.0193 Inf 0.1763 0.252
## BDCT High production 0.220 0.0292 Inf 0.1625 0.277
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
tab$outcome <- "IMI at calving"
tab.imipc <- tab
tab.imipc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.05 (-0.02 to 0.12) IMI at calving
## 2 Low production A_SDCT - BDCT 0 (-0.04 to 0.03) IMI at calving
## 3 Mid production C_SDCT - BDCT 0 (-0.05 to 0.05) IMI at calving
## 4 Mid production A_SDCT - BDCT 0.01 (-0.04 to 0.07) IMI at calving
## 5 High production C_SDCT - BDCT 0.01 (-0.02 to 0.05) IMI at calving
## 6 High production A_SDCT - BDCT -0.01 (-0.06 to 0.05) IMI at calvingtab.imipc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.05 (-0.02 to 0.12) IMI at calving
## 2 Low production A_SDCT - BDCT 0 (-0.04 to 0.03) IMI at calving
## 3 Mid production C_SDCT - BDCT 0 (-0.05 to 0.05) IMI at calving
## 4 Mid production A_SDCT - BDCT 0.01 (-0.04 to 0.07) IMI at calving
## 5 High production C_SDCT - BDCT 0.01 (-0.02 to 0.05) IMI at calving
## 6 High production A_SDCT - BDCT -0.01 (-0.06 to 0.05) IMI at calvingUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(imipc ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(imipc ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: imipc ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 5106 5198 -2539 5078 5079
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -0.906 -0.598 -0.436 -0.239 4.829
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.392 0.626
## Number of obs: 5093, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.9070 0.2908 -6.56 5.5e-11 ***
## TxA_SDCT -0.3099 0.1638 -1.89 0.0585 .
## TxBDCT -0.2718 0.1395 -1.95 0.0513 .
## my3Mid production -0.2117 0.1473 -1.44 0.1505
## my3High production -0.1097 0.1514 -0.72 0.4686
## Parity2 0.1638 0.0868 1.89 0.0593 .
## Parity3 0.2391 0.0895 2.67 0.0075 **
## DOSCC 0.0810 0.0335 2.42 0.0156 *
## factor(source)US -0.1493 0.3382 -0.44 0.6590
## TxA_SDCT:my3Mid production 0.4109 0.2234 1.84 0.0658 .
## TxBDCT:my3Mid production 0.2618 0.2006 1.31 0.1919
## TxA_SDCT:my3High production 0.2094 0.2190 0.96 0.3390
## TxBDCT:my3High production 0.2636 0.1991 1.32 0.1854
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: imipc
## Chisq Df Pr(>Chisq)
## Tx 1.86 2 0.395
## my3 0.30 2 0.862
## Parity 7.70 2 0.021 *
## DOSCC 5.84 1 0.016 *
## factor(source) 0.19 1 0.659
## Tx:my3 4.45 4 0.348
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: imipc ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 5288
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.0328 -0.6519 -0.4101 -0.0958 2.4775
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.0078 0.0883
## Residual 0.1620 0.4025
## Number of obs: 5093, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.13198 0.04381 3.01
## TxA_SDCT -0.05075 0.02669 -1.90
## TxBDCT -0.04369 0.02260 -1.93
## my3Mid production -0.03508 0.02405 -1.46
## my3High production -0.01799 0.02515 -0.72
## Parity2 0.02529 0.01412 1.79
## Parity3 0.03795 0.01471 2.58
## DOSCC 0.01320 0.00555 2.38
## factor(source)US -0.00558 0.04782 -0.12
## TxA_SDCT:my3Mid production 0.06646 0.03636 1.83
## TxBDCT:my3Mid production 0.04117 0.03217 1.28
## TxA_SDCT:my3High production 0.03316 0.03585 0.93
## TxBDCT:my3High production 0.04203 0.03258 1.29car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: imipc
## Chisq Df Pr(>Chisq)
## Tx 1.91 2 0.385
## my3 0.35 2 0.840
## Parity 7.19 2 0.028 *
## DOSCC 5.65 1 0.017 *
## factor(source) 0.01 1 0.907
## Tx:my3 4.33 4 0.363
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.184 0.0289 Inf 0.1340 0.247
## A_SDCT Low production 0.142 0.0258 Inf 0.0985 0.200
## BDCT Low production 0.147 0.0239 Inf 0.1057 0.200
## C_SDCT Mid production 0.154 0.0253 Inf 0.1110 0.211
## A_SDCT Mid production 0.168 0.0287 Inf 0.1189 0.232
## BDCT Mid production 0.153 0.0251 Inf 0.1101 0.209
## C_SDCT High production 0.168 0.0273 Inf 0.1211 0.229
## A_SDCT High production 0.155 0.0266 Inf 0.1092 0.214
## BDCT High production 0.167 0.0275 Inf 0.1199 0.228
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.04 (0 to 0.09)
## 2 Low production A_SDCT - BDCT -0.01 (-0.06 to 0.04)
## 3 Mid production C_SDCT - BDCT 0 (-0.04 to 0.05)
## 4 Mid production A_SDCT - BDCT 0.02 (-0.03 to 0.07)
## 5 High production C_SDCT - BDCT 0 (-0.04 to 0.05)
## 6 High production A_SDCT - BDCT -0.02 (-0.06 to 0.03)IMI cure
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(cure ~ Tx + my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
imicure.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
imicure.main <- imicure.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
imicure.main$production <- "All production levels"
imicure.main$contrast <- "none"
imicure.main$outcome <- "IMI cure"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
gee <- geeglm(cure ~ Tx*my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ Tx * my3 + Parity + DOSCC + factor(source),
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(cure) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 1.00675 0.03928 657.00 <2e-16 ***
## TxA_SDCT 0.01459 0.02525 0.33 0.563
## TxBDCT -0.03046 0.02715 1.26 0.262
## my3Mid production -0.02144 0.02231 0.92 0.337
## my3High production 0.02009 0.02314 0.75 0.385
## Parity2 -0.03668 0.01403 6.83 0.009 **
## Parity3 -0.04331 0.03168 1.87 0.172
## DOSCC -0.01202 0.00617 3.80 0.051 .
## factor(source)US -0.05330 0.03086 2.98 0.084 .
## TxA_SDCT:my3Mid production 0.02096 0.03195 0.43 0.512
## TxBDCT:my3Mid production 0.05894 0.02763 4.55 0.033 *
## TxA_SDCT:my3High production -0.03436 0.04932 0.49 0.486
## TxBDCT:my3High production -0.02447 0.03060 0.64 0.424
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.287
## Number of clusters: 16 Maximum cluster size: 432glm <- glm(cure ~ alg*my3 + cult*my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"))
int.p <- car::Anova(glm) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "imicure"
int.p.cure <- int.p
emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.898 0.0264 Inf 0.846 0.950
## A_SDCT Low production 0.913 0.0324 Inf 0.849 0.976
## BDCT Low production 0.868 0.0389 Inf 0.791 0.944
## C_SDCT Mid production 0.877 0.0238 Inf 0.830 0.923
## A_SDCT Mid production 0.912 0.0143 Inf 0.884 0.940
## BDCT Mid production 0.905 0.0172 Inf 0.871 0.939
## C_SDCT High production 0.918 0.0131 Inf 0.893 0.944
## A_SDCT High production 0.898 0.0258 Inf 0.848 0.949
## BDCT High production 0.863 0.0163 Inf 0.831 0.895
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
tab$outcome <- "IMI cure"
tab.imicure <- tab
tab.imicure## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.03 (-0.02 to 0.08) IMI cure
## 2 Low production A_SDCT - BDCT 0.05 (-0.02 to 0.11) IMI cure
## 3 Mid production C_SDCT - BDCT -0.03 (-0.09 to 0.03) IMI cure
## 4 Mid production A_SDCT - BDCT 0.01 (-0.02 to 0.03) IMI cure
## 5 High production C_SDCT - BDCT 0.05 (0.02 to 0.09) IMI cure
## 6 High production A_SDCT - BDCT 0.04 (-0.02 to 0.09) IMI cureUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(cure ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(cure ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: cure ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 875 946 -424 847 1171
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.961 0.260 0.333 0.401 0.582
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.328 0.573
## Number of obs: 1185, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 2.8963 0.6433 4.50 6.7e-06 ***
## TxA_SDCT 0.1204 0.4931 0.24 0.807
## TxBDCT -0.1422 0.3834 -0.37 0.711
## my3Mid production -0.1459 0.4015 -0.36 0.716
## my3High production 0.3537 0.4223 0.84 0.402
## Parity2 -0.3767 0.2213 -1.70 0.089 .
## Parity3 -0.2253 0.2349 -0.96 0.337
## DOSCC -0.0637 0.0912 -0.70 0.485
## factor(source)US -0.1004 0.4187 -0.24 0.810
## TxA_SDCT:my3Mid production -0.0131 0.6254 -0.02 0.983
## TxBDCT:my3Mid production 0.5154 0.5397 0.96 0.340
## TxA_SDCT:my3High production -0.2580 0.6272 -0.41 0.681
## TxBDCT:my3High production -0.3065 0.5244 -0.58 0.559
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## optimizer (Nelder_Mead) convergence code: 0 (OK)
## Model failed to converge with max|grad| = 0.00282032 (tol = 0.002, component 1)car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: cure
## Chisq Df Pr(>Chisq)
## Tx 0.39 2 0.82
## my3 0.65 2 0.72
## Parity 2.93 2 0.23
## DOSCC 0.49 1 0.48
## factor(source) 0.06 1 0.81
## Tx:my3 2.70 4 0.61summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: cure ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 748
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.949 0.201 0.326 0.446 0.688
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.00297 0.0545
## Residual 0.10365 0.3219
## Number of obs: 1185, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.95152 0.06273 15.17
## TxA_SDCT 0.01292 0.05078 0.25
## TxBDCT -0.01566 0.04005 -0.39
## my3Mid production -0.01576 0.04185 -0.38
## my3High production 0.03373 0.04168 0.81
## Parity2 -0.03796 0.02299 -1.65
## Parity3 -0.02184 0.02410 -0.91
## DOSCC -0.00613 0.00936 -0.65
## factor(source)US -0.00602 0.03895 -0.15
## TxA_SDCT:my3Mid production -0.00220 0.06520 -0.03
## TxBDCT:my3Mid production 0.05305 0.05537 0.96
## TxA_SDCT:my3High production -0.02312 0.06308 -0.37
## TxBDCT:my3High production -0.02941 0.05416 -0.54car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: cure
## Chisq Df Pr(>Chisq)
## Tx 0.32 2 0.85
## my3 0.70 2 0.70
## Parity 2.78 2 0.25
## DOSCC 0.43 1 0.51
## factor(source) 0.02 1 0.88
## Tx:my3 2.74 4 0.60emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.913 0.0287 Inf 0.838 0.955
## A_SDCT Low production 0.922 0.0320 Inf 0.832 0.966
## BDCT Low production 0.901 0.0278 Inf 0.832 0.944
## C_SDCT Mid production 0.901 0.0286 Inf 0.829 0.945
## A_SDCT Mid production 0.910 0.0309 Inf 0.828 0.955
## BDCT Mid production 0.930 0.0234 Inf 0.868 0.964
## C_SDCT High production 0.937 0.0206 Inf 0.883 0.968
## A_SDCT High production 0.929 0.0245 Inf 0.863 0.964
## BDCT High production 0.905 0.0281 Inf 0.834 0.948
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["lower.CL"]],
ucl = out[["contrasts"]][["upper.CL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.02 (-0.06 to 0.09)
## 2 Low production A_SDCT - BDCT 0.03 (-0.07 to 0.12)
## 3 Mid production C_SDCT - BDCT -0.04 (-0.11 to 0.04)
## 4 Mid production A_SDCT - BDCT -0.03 (-0.11 to 0.06)
## 5 High production C_SDCT - BDCT 0.05 (-0.03 to 0.12)
## 6 High production A_SDCT - BDCT 0.03 (-0.04 to 0.11)NewIMI
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
gee <- geeglm(newimi ~ Tx + my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
iminew.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
iminew.main <- iminew.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
iminew.main$production <- "All production levels"
iminew.main$contrast <- "none"
iminew.main$outcome <- "New IMI risk"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
gee <- geeglm(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) , data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ Tx * my3 + Parity + DOSCC + factor(source),
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(newimi) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.05905 0.06482 0.83 0.3623
## TxA_SDCT -0.04170 0.03348 1.55 0.2130
## TxBDCT -0.05209 0.04118 1.60 0.2059
## my3Mid production -0.03708 0.02879 1.66 0.1978
## my3High production 0.01131 0.03407 0.11 0.7399
## Parity2 0.00815 0.02343 0.12 0.7279
## Parity3 0.01569 0.01766 0.79 0.3742
## DOSCC 0.02704 0.00925 8.54 0.0035 **
## factor(source)US 0.05721 0.04858 1.39 0.2389
## TxA_SDCT:my3Mid production 0.05020 0.04223 1.41 0.2345
## TxBDCT:my3Mid production 0.05598 0.03817 2.15 0.1425
## TxA_SDCT:my3High production 0.01119 0.04701 0.06 0.8119
## TxBDCT:my3High production 0.02649 0.03685 0.52 0.4722
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.17
## Number of clusters: 16 Maximum cluster size: 1223glm <- glm(newimi ~ alg*my3 + cult*my3 + Parity + DOSCC + factor(source) , data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"))
int.p <- car::Anova(glm) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "newimi"
int.p.newimi <- int.p
emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.216 0.0308 Inf 0.1555 0.276
## A_SDCT Low production 0.174 0.0327 Inf 0.1100 0.238
## BDCT Low production 0.164 0.0393 Inf 0.0866 0.241
## C_SDCT Mid production 0.179 0.0271 Inf 0.1256 0.232
## A_SDCT Mid production 0.187 0.0260 Inf 0.1363 0.238
## BDCT Mid production 0.183 0.0351 Inf 0.1139 0.251
## C_SDCT High production 0.227 0.0223 Inf 0.1835 0.271
## A_SDCT High production 0.197 0.0189 Inf 0.1595 0.234
## BDCT High production 0.202 0.0330 Inf 0.1368 0.266
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
#mean
tab$outcome <- "New IMI risk"
tab.newimi <- tab
tab.newimi## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.05 (-0.03 to 0.13) New IMI risk
## 2 Low production A_SDCT - BDCT 0.01 (-0.02 to 0.04) New IMI risk
## 3 Mid production C_SDCT - BDCT 0 (-0.05 to 0.04) New IMI risk
## 4 Mid production A_SDCT - BDCT 0 (-0.04 to 0.05) New IMI risk
## 5 High production C_SDCT - BDCT 0.03 (-0.02 to 0.07) New IMI risk
## 6 High production A_SDCT - BDCT 0 (-0.07 to 0.06) New IMI riskUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: newimi ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 4481 4572 -2227 4453 4620
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -0.915 -0.591 -0.401 -0.233 4.996
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.423 0.65
## Number of obs: 4634, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.8922 0.3033 -6.24 4.4e-10 ***
## TxA_SDCT -0.2704 0.1710 -1.58 0.114
## TxBDCT -0.3038 0.1466 -2.07 0.038 *
## my3Mid production -0.3388 0.1573 -2.15 0.031 *
## my3High production -0.1451 0.1611 -0.90 0.368
## Parity2 0.1988 0.0929 2.14 0.032 *
## Parity3 0.2634 0.0958 2.75 0.006 **
## DOSCC 0.0696 0.0353 1.97 0.048 *
## factor(source)US -0.1715 0.3522 -0.49 0.626
## TxA_SDCT:my3Mid production 0.3685 0.2383 1.55 0.122
## TxBDCT:my3Mid production 0.3504 0.2141 1.64 0.102
## TxA_SDCT:my3High production 0.1042 0.2321 0.45 0.654
## TxBDCT:my3High production 0.2214 0.2130 1.04 0.299
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: newimi
## Chisq Df Pr(>Chisq)
## Tx 2.37 2 0.305
## my3 1.38 2 0.501
## Parity 8.58 2 0.014 *
## DOSCC 3.90 1 0.048 *
## factor(source) 0.24 1 0.626
## Tx:my3 3.93 4 0.415
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: newimi ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 4607
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.051 -0.659 -0.367 -0.106 2.528
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.00792 0.089
## Residual 0.15472 0.393
## Number of obs: 4634, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.13930 0.04413 3.16
## TxA_SDCT -0.04298 0.02719 -1.58
## TxBDCT -0.04737 0.02289 -2.07
## my3Mid production -0.05305 0.02446 -2.17
## my3High production -0.02283 0.02576 -0.89
## Parity2 0.02945 0.01444 2.04
## Parity3 0.04016 0.01505 2.67
## DOSCC 0.01076 0.00560 1.92
## factor(source)US -0.00675 0.04816 -0.14
## TxA_SDCT:my3Mid production 0.05684 0.03716 1.53
## TxBDCT:my3Mid production 0.05315 0.03273 1.62
## TxA_SDCT:my3High production 0.01513 0.03677 0.41
## TxBDCT:my3High production 0.03397 0.03342 1.02car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: newimi
## Chisq Df Pr(>Chisq)
## Tx 2.39 2 0.302
## my3 1.52 2 0.469
## Parity 8.07 2 0.018 *
## DOSCC 3.68 1 0.055 .
## factor(source) 0.02 1 0.889
## Tx:my3 3.90 4 0.420
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.180 0.0296 Inf 0.1294 0.246
## A_SDCT Low production 0.144 0.0271 Inf 0.0983 0.205
## BDCT Low production 0.140 0.0240 Inf 0.0989 0.194
## C_SDCT Mid production 0.136 0.0239 Inf 0.0950 0.190
## A_SDCT Mid production 0.147 0.0274 Inf 0.1014 0.209
## BDCT Mid production 0.141 0.0246 Inf 0.0994 0.196
## C_SDCT High production 0.160 0.0276 Inf 0.1129 0.222
## A_SDCT High production 0.139 0.0257 Inf 0.0956 0.197
## BDCT High production 0.149 0.0265 Inf 0.1042 0.209
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.05 (0 to 0.09)
## 2 Low production A_SDCT - BDCT 0 (-0.05 to 0.06)
## 3 Mid production C_SDCT - BDCT -0.01 (-0.05 to 0.04)
## 4 Mid production A_SDCT - BDCT 0.01 (-0.04 to 0.06)
## 5 High production C_SDCT - BDCT 0.01 (-0.03 to 0.06)
## 6 High production A_SDCT - BDCT -0.01 (-0.07 to 0.04)DHI Dataset
Somatic cell count (1-120 DIM)
All production groups pooled together
dhi$Tx <- fct_relevel(dhi$Tx,"BDCT","C_SDCT","A_SDCT")
dhi <- dhi %>% arrange(FARMID,CowID)
mm0 <- geeglm(scc ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID),
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID), data = dhi %>% filter(!is.na(DOSCC)), id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.3045 0.1784 343.27 < 2e-16 ***
## TxC_SDCT 0.0278 0.0639 0.19 0.6631
## TxA_SDCT 0.0273 0.0684 0.16 0.6903
## my3Mid production 0.0421 0.0720 0.34 0.5589
## my3High production 0.0755 0.0763 0.98 0.3219
## TestDIMcat2030 -0.5416 0.0646 70.37 < 2e-16 ***
## TestDIMcat2050 -0.5565 0.0589 89.18 < 2e-16 ***
## TestDIMcat2070 -0.4182 0.0606 47.60 5.2e-12 ***
## TestDIMcat2090 -0.2862 0.0648 19.51 1.0e-05 ***
## TestDIMcat20110 -0.1244 0.0611 4.15 0.0417 *
## DOSCC 0.1759 0.0268 43.02 5.4e-11 ***
## Parity2 0.1333 0.0625 4.55 0.0329 *
## Parity3 0.3022 0.0735 16.90 3.9e-05 ***
## factor(FARMID)2 0.0545 0.1709 0.10 0.7497
## factor(FARMID)3 -0.0429 0.1346 0.10 0.7499
## factor(FARMID)4 0.0336 0.1316 0.07 0.7981
## factor(FARMID)5 0.0314 0.1357 0.05 0.8171
## factor(FARMID)6 -0.5076 0.1803 7.92 0.0049 **
## factor(FARMID)7 -0.3009 0.1492 4.07 0.0437 *
## factor(FARMID)8 -0.0673 0.2795 0.06 0.8097
## factor(FARMID)9 -0.0525 0.2980 0.03 0.8602
## factor(FARMID)10 -0.2642 0.2269 1.36 0.2443
## factor(FARMID)11 -0.4061 0.2410 2.84 0.0919 .
## factor(FARMID)12 -0.4632 0.2170 4.56 0.0328 *
## factor(FARMID)13 0.1282 0.3030 0.18 0.6722
## factor(FARMID)14 0.0986 0.2545 0.15 0.6983
## factor(FARMID)15 0.0221 0.3689 0.00 0.9521
## factor(FARMID)16 -0.0613 0.2328 0.07 0.7923
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0566
## Number of clusters: 1364 Maximum cluster size: 4ggplot(data.frame(eta=predict(mm0,type="link"),pearson=residuals(mm0,type="pearson")),
aes(x=eta,y=pearson)) + geom_point() + theme_bw()
qqnorm(residuals(mm0))
scc.main <- tidy(mm0,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
scc.main <- scc.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
scc.main$production <- "All production levels"
scc.main$contrast <- "none"
scc.main$outcome <- "Log SCC (1-120 DIM)"atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(Tx,TestDIMcat20,SCC,LCL,UCL))
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
emm## # A tibble: 18 × 5
## Tx TestDIMcat20 SCC LCL UCL
## <fct> <chr> <dbl> <dbl> <dbl>
## 1 BDCT 10 63.6 55.3 73.1
## 2 C_SDCT 10 65.4 56.8 75.3
## 3 A_SDCT 10 65.4 55.5 76.9
## 4 BDCT 30 37.0 32.1 42.6
## 5 C_SDCT 30 38.0 33.0 43.8
## 6 A_SDCT 30 38.0 32.4 44.7
## 7 BDCT 50 36.5 31.8 41.8
## 8 C_SDCT 50 37.5 32.6 43.2
## 9 A_SDCT 50 37.5 31.9 44.0
## 10 BDCT 70 41.9 36.3 48.3
## 11 C_SDCT 70 43.0 37.5 49.5
## 12 A_SDCT 70 43.0 36.6 50.5
## 13 BDCT 90 47.8 41.5 55.0
## 14 C_SDCT 90 49.1 42.4 56.8
## 15 A_SDCT 90 49.1 41.7 57.8
## 16 BDCT 110 56.2 48.8 64.6
## 17 C_SDCT 110 57.7 50.1 66.6
## 18 A_SDCT 110 57.7 49.1 67.9curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=Tx, colour=Tx) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$LCL, ymax=emm$UCL,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve
Interaction model (Tx * milk yield)
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(scc ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID),
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID), data = dhi %>% filter(!is.na(DOSCC)), id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.2661 0.1929 286.67 < 2e-16 ***
## TxA_SDCT 0.1569 0.1267 1.53 0.2155
## TxBDCT 0.0084 0.1101 0.01 0.9392
## my3Mid production 0.0828 0.1141 0.53 0.4683
## my3High production 0.2013 0.1197 2.83 0.0928 .
## TestDIMcat2030 -0.5401 0.0645 70.06 < 2e-16 ***
## TestDIMcat2050 -0.5556 0.0589 88.90 < 2e-16 ***
## TestDIMcat2070 -0.4169 0.0606 47.35 5.9e-12 ***
## TestDIMcat2090 -0.2846 0.0648 19.30 1.1e-05 ***
## TestDIMcat20110 -0.1225 0.0611 4.02 0.0449 *
## DOSCC 0.1781 0.0269 43.89 3.5e-11 ***
## Parity2 0.1315 0.0627 4.40 0.0359 *
## Parity3 0.3010 0.0732 16.94 3.9e-05 ***
## factor(FARMID)2 0.0517 0.1726 0.09 0.7647
## factor(FARMID)3 -0.0427 0.1354 0.10 0.7525
## factor(FARMID)4 0.0285 0.1327 0.05 0.8298
## factor(FARMID)5 0.0224 0.1361 0.03 0.8691
## factor(FARMID)6 -0.5184 0.1810 8.20 0.0042 **
## factor(FARMID)7 -0.3022 0.1498 4.07 0.0437 *
## factor(FARMID)8 -0.0442 0.2805 0.02 0.8747
## factor(FARMID)9 -0.0294 0.3004 0.01 0.9220
## factor(FARMID)10 -0.2859 0.2294 1.55 0.2128
## factor(FARMID)11 -0.3844 0.2408 2.55 0.1104
## factor(FARMID)12 -0.4512 0.2202 4.20 0.0405 *
## factor(FARMID)13 0.1514 0.3046 0.25 0.6193
## factor(FARMID)14 0.1086 0.2497 0.19 0.6637
## factor(FARMID)15 0.0274 0.3643 0.01 0.9400
## factor(FARMID)16 -0.0574 0.2329 0.06 0.8053
## TxA_SDCT:my3Mid production -0.1910 0.1708 1.25 0.2634
## TxBDCT:my3Mid production 0.0425 0.1533 0.08 0.7816
## TxA_SDCT:my3High production -0.2570 0.1686 2.32 0.1273
## TxBDCT:my3High production -0.1480 0.1573 0.89 0.3466
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0562
## Number of clusters: 1364 Maximum cluster size: 4lmer <- lmer(scc ~ alg*my3 + cult*my3 + TestDIMcat20 + DOSCC + Parity + (1|FARMID/CowID),
data=dhi %>% filter(!is.na(DOSCC)))
int.p <- car::Anova(lmer) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "scc"
int.p.scc <- int.p
out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Log SCC (1-120 DIM)"
tab.scc <- tab
tab.scc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT -0.01 (-0.22 to 0.21) Log SCC (1-120 DIM)
## 2 Low production A_SDCT - BDCT 0.15 (-0.09 to 0.39) Log SCC (1-120 DIM)
## 3 Mid production C_SDCT - BDCT -0.05 (-0.26 to 0.16) Log SCC (1-120 DIM)
## 4 Mid production A_SDCT - BDCT -0.09 (-0.31 to 0.14) Log SCC (1-120 DIM)
## 5 High production C_SDCT - BDCT 0.14 (-0.08 to 0.36) Log SCC (1-120 DIM)
## 6 High production A_SDCT - BDCT 0.04 (-0.19 to 0.26) Log SCC (1-120 DIM)atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(my3,Tx,TestDIMcat20,SCC,LCL,UCL))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=Tx, colour=Tx) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=LCL, ymax=UCL,group=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",
legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1") + facet_grid( ~ my3)
tiff("curve1.tiff", units="in", width=10, height=6, res=300)
curve
dev.off()## quartz_off_screen
## 2curve
Interaction model using linear mixed model
lmm <- lmer(scc ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + (1|FARMID/CowID),
data=dhi)
summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: scc ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity + (1 | FARMID/CowID)
## Data: dhi
##
## REML criterion at convergence: 14784
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.136 -0.523 -0.110 0.436 4.116
##
## Random effects:
## Groups Name Variance Std.Dev.
## CowID:FARMID (Intercept) 0.701 0.837
## FARMID (Intercept) 0.014 0.118
## Residual 1.053 1.026
## Number of obs: 4553, groups: CowID:FARMID, 1364; FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 3.16592 0.15439 20.51
## TxA_SDCT 0.17014 0.12451 1.37
## TxBDCT 0.00547 0.11058 0.05
## my3Mid production 0.06995 0.11663 0.60
## my3High production 0.19372 0.11878 1.63
## TestDIMcat2030 -0.56693 0.05762 -9.84
## TestDIMcat2050 -0.57328 0.05519 -10.39
## TestDIMcat2070 -0.41919 0.05476 -7.65
## TestDIMcat2090 -0.29961 0.05865 -5.11
## TestDIMcat20110 -0.11768 0.05479 -2.15
## DOSCC 0.18835 0.02625 7.18
## Parity2 0.10068 0.06828 1.47
## Parity3 0.29597 0.07124 4.15
## TxA_SDCT:my3Mid production -0.14597 0.17226 -0.85
## TxBDCT:my3Mid production 0.02985 0.15757 0.19
## TxA_SDCT:my3High production -0.25644 0.17125 -1.50
## TxBDCT:my3High production -0.13304 0.16038 -0.83car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: scc
## Chisq Df Pr(>Chisq)
## Tx 0.75 2 0.68885
## my3 0.98 2 0.61214
## TestDIMcat20 183.45 5 < 2e-16 ***
## DOSCC 51.50 1 7.2e-13 ***
## Parity 17.41 2 0.00017 ***
## Tx:my3 2.99 4 0.55912
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT -0.01 (-0.22 to 0.21)
## 2 Low production A_SDCT - BDCT 0.16 (-0.07 to 0.4)
## 3 Mid production C_SDCT - BDCT -0.04 (-0.26 to 0.18)
## 4 Mid production A_SDCT - BDCT -0.01 (-0.25 to 0.22)
## 5 High production C_SDCT - BDCT 0.13 (-0.1 to 0.36)
## 6 High production A_SDCT - BDCT 0.04 (-0.2 to 0.28)Milk Yield (1-120 DIM)
All production groups pooled together
dhi$Tx <- fct_relevel(dhi$Tx,"BDCT","C_SDCT","A_SDCT")
dhi <- dhi %>% arrange(FARMID,CowID)
mm0 <- geeglm(MY ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID) + DOMY,
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID) + DOMY, data = dhi %>% filter(!is.na(DOSCC)),
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 31.8031 1.5941 398.04 < 2e-16 ***
## TxC_SDCT -0.1808 0.4107 0.19 0.65965
## TxA_SDCT -0.9814 0.4546 4.66 0.03084 *
## my3Mid production -0.5712 0.6640 0.74 0.38971
## my3High production 0.5849 0.9623 0.37 0.54328
## TestDIMcat2030 10.6606 0.4583 541.14 < 2e-16 ***
## TestDIMcat2050 12.3200 0.3998 949.77 < 2e-16 ***
## TestDIMcat2070 11.5855 0.4275 734.54 < 2e-16 ***
## TestDIMcat2090 10.1087 0.4414 524.56 < 2e-16 ***
## TestDIMcat20110 7.9665 0.4369 332.43 < 2e-16 ***
## DOSCC 0.1355 0.1756 0.59 0.44052
## Parity2 2.8994 0.4220 47.21 6.4e-12 ***
## Parity3 3.3581 0.4942 46.16 1.1e-11 ***
## factor(FARMID)2 -1.6117 1.0459 2.37 0.12330
## factor(FARMID)3 -4.9980 0.8516 34.44 4.4e-09 ***
## factor(FARMID)4 4.8781 0.8744 31.12 2.4e-08 ***
## factor(FARMID)5 0.9036 0.8896 1.03 0.30974
## factor(FARMID)6 1.8089 1.0391 3.03 0.08172 .
## factor(FARMID)7 1.6084 0.9903 2.64 0.10434
## factor(FARMID)8 -1.8911 1.6306 1.35 0.24615
## factor(FARMID)9 -7.7980 1.6977 21.10 4.4e-06 ***
## factor(FARMID)10 -1.6553 1.2098 1.87 0.17125
## factor(FARMID)11 -3.2094 1.7474 3.37 0.06625 .
## factor(FARMID)12 -4.1596 1.7436 5.69 0.01705 *
## factor(FARMID)13 -4.8266 1.2447 15.04 0.00011 ***
## factor(FARMID)14 -2.0698 2.4659 0.70 0.40127
## factor(FARMID)15 -4.9931 1.6537 9.12 0.00253 **
## factor(FARMID)16 -2.3555 1.2312 3.66 0.05574 .
## DOMY 0.1937 0.0484 15.99 6.4e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 71.3 2.37
## Number of clusters: 1363 Maximum cluster size: 4ggplot(data.frame(eta=predict(mm0,type="link"),pearson=residuals(mm0,type="pearson")),
aes(x=eta,y=pearson)) + geom_point() + theme_bw()
qqnorm(residuals(mm0))
my.main <- tidy(mm0,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
my.main <- my.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
my.main$production <- "All production levels"
my.main$contrast <- "none"
my.main$outcome <- "Milk Yield (1-120 DIM)"atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm <- emm %>% subset(select=c(Tx,TestDIMcat20,estimate,conf.low,conf.high))
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(30,55))) + aes(x=as.numeric(TestDIMcat20), y=estimate, group=Tx, colour=Tx) +
labs(y = "Milk Yield)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$conf.low, ymax=emm$conf.high,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve
Interaction model (Tx * milk yield)
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID) + DOMY,
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID) + DOMY, data = dhi %>% filter(!is.na(DOSCC)),
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 31.9459 1.6143 391.62 < 2e-16 ***
## TxA_SDCT -1.6111 0.8959 3.23 0.0721 .
## TxBDCT 0.0852 0.6959 0.01 0.9026
## my3Mid production -0.7481 0.8719 0.74 0.3909
## my3High production 0.0409 1.1758 0.00 0.9722
## TestDIMcat2030 10.6529 0.4586 539.63 < 2e-16 ***
## TestDIMcat2050 12.3127 0.3997 948.81 < 2e-16 ***
## TestDIMcat2070 11.5762 0.4274 733.67 < 2e-16 ***
## TestDIMcat2090 10.0970 0.4416 522.79 < 2e-16 ***
## TestDIMcat20110 7.9542 0.4368 331.58 < 2e-16 ***
## DOSCC 0.1279 0.1756 0.53 0.4664
## Parity2 2.8962 0.4200 47.55 5.4e-12 ***
## Parity3 3.3503 0.4930 46.18 1.1e-11 ***
## factor(FARMID)2 -1.6389 1.0480 2.45 0.1178
## factor(FARMID)3 -5.0154 0.8553 34.38 4.5e-09 ***
## factor(FARMID)4 4.8965 0.8796 30.99 2.6e-08 ***
## factor(FARMID)5 0.9276 0.8942 1.08 0.2996
## factor(FARMID)6 1.8861 1.0472 3.24 0.0717 .
## factor(FARMID)7 1.6251 0.9947 2.67 0.1023
## factor(FARMID)8 -2.0086 1.6474 1.49 0.2228
## factor(FARMID)9 -7.8925 1.7156 21.16 4.2e-06 ***
## factor(FARMID)10 -1.4838 1.2213 1.48 0.2244
## factor(FARMID)11 -3.3325 1.7722 3.54 0.0601 .
## factor(FARMID)12 -4.1916 1.7713 5.60 0.0180 *
## factor(FARMID)13 -4.9395 1.2652 15.24 9.5e-05 ***
## factor(FARMID)14 -2.1162 2.4571 0.74 0.3891
## factor(FARMID)15 -5.0086 1.6358 9.37 0.0022 **
## factor(FARMID)16 -2.3548 1.2507 3.54 0.0597 .
## DOMY 0.1925 0.0485 15.74 7.3e-05 ***
## TxA_SDCT:my3Mid production 1.3455 1.1085 1.47 0.2248
## TxBDCT:my3Mid production -0.5500 0.9506 0.33 0.5629
## TxA_SDCT:my3High production 1.0020 1.2024 0.69 0.4047
## TxBDCT:my3High production 0.8429 1.0448 0.65 0.4198
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 71.1 2.36
## Number of clusters: 1363 Maximum cluster size: 4lmer <- lmer(scc ~ my3*cult + my3*alg+ TestDIMcat20 + DOSCC + Parity + DOMY + (1|FARMID/CowID),
data=dhi %>% filter(!is.na(DOSCC)))
int.p <- car::Anova(lmer) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "my"
int.p.my <- int.p
out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Milk Yield"
tab.my <- tab
tab.my## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT -0.09 (-1.45 to 1.28) Milk Yield
## 2 Low production A_SDCT - BDCT -1.7 (-3.39 to 0) Milk Yield
## 3 Mid production C_SDCT - BDCT 0.46 (-0.82 to 1.75) Milk Yield
## 4 Mid production A_SDCT - BDCT 0.2 (-1.11 to 1.5) Milk Yield
## 5 High production C_SDCT - BDCT -0.93 (-2.45 to 0.59) Milk Yield
## 6 High production A_SDCT - BDCT -1.54 (-3.11 to 0.04) Milk Yieldatx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm <- emm %>% subset(select=c(my3,Tx,TestDIMcat20,estimate,conf.low,conf.high))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(30,55))) + aes(x=as.numeric(TestDIMcat20), y=estimate, group=Tx, colour=Tx) +
labs(y = "Milk Yield (kg/d)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$conf.low, ymax=emm$conf.high,colour=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1") + facet_grid( ~ my3)
tiff("curve2.tiff", units="in", width=10, height=6, res=300)
curve
dev.off() ## quartz_off_screen
## 2curve
Interaction model using linear mixed model
lmm <- lmer(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + (1|FARMID/CowID),
data=dhi)
summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity + (1 | FARMID/CowID)
## Data: dhi
##
## REML criterion at convergence: 31700
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.498 -0.468 0.021 0.518 4.182
##
## Random effects:
## Groups Name Variance Std.Dev.
## CowID:FARMID (Intercept) 29.14 5.40
## FARMID (Intercept) 9.33 3.06
## Residual 43.72 6.61
## Number of obs: 4551, groups: CowID:FARMID, 1363; FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 34.0021 1.2497 27.21
## TxA_SDCT -1.3972 0.8145 -1.72
## TxBDCT 0.0731 0.7140 0.10
## my3Mid production 1.2098 0.7631 1.59
## my3High production 3.5547 0.7856 4.52
## TestDIMcat2030 11.1954 0.3718 30.11
## TestDIMcat2050 12.2545 0.3558 34.44
## TestDIMcat2070 11.8824 0.3531 33.65
## TestDIMcat2090 10.2685 0.3782 27.15
## TestDIMcat20110 7.9111 0.3533 22.39
## DOSCC -0.0128 0.1723 -0.07
## Parity2 2.9730 0.4420 6.73
## Parity3 3.1969 0.4621 6.92
## TxA_SDCT:my3Mid production 1.1928 1.1170 1.07
## TxBDCT:my3Mid production -0.5866 1.0179 -0.58
## TxA_SDCT:my3High production 0.8122 1.1118 0.73
## TxBDCT:my3High production 0.8237 1.0375 0.79car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: MY
## Chisq Df Pr(>Chisq)
## Tx 3.81 2 0.15
## my3 69.16 2 9.6e-16 ***
## TestDIMcat20 1659.54 5 < 2e-16 ***
## DOSCC 0.01 1 0.94
## Parity 64.68 2 9.0e-15 ***
## Tx:my3 4.33 4 0.36
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1int.p <- car::Anova(lmer) %>% slice_tail(n=2) %>% select(`Pr(>Chisq)`)
int.p <- tibble::rownames_to_column(int.p, "interaction")
int.p$outcome <- "scc"
int.p.scc <- int.p
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
icc(lmm)## # Intraclass Correlation Coefficient
##
## Adjusted ICC: 0.468
## Unadjusted ICC: 0.367tab## production contrast estimate
## 1 Low production C_SDCT - BDCT -0.07 (-1.47 to 1.33)
## 2 Low production A_SDCT - BDCT -1.47 (-3.01 to 0.07)
## 3 Mid production C_SDCT - BDCT 0.51 (-0.91 to 1.94)
## 4 Mid production A_SDCT - BDCT 0.31 (-1.21 to 1.83)
## 5 High production C_SDCT - BDCT -0.9 (-2.37 to 0.58)
## 6 High production A_SDCT - BDCT -1.48 (-3.03 to 0.07)Performing analysis without including farmID as fixed or random effect (milk yield)
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + DOMY,
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## DOMY, data = dhi %>% filter(!is.na(DOSCC)), id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 28.4835 1.5742 327.40 < 2e-16 ***
## TxA_SDCT -0.1063 0.9654 0.01 0.912
## TxBDCT 0.4428 0.7748 0.33 0.568
## my3Mid production 0.4428 0.9665 0.21 0.647
## my3High production 2.6525 1.2674 4.38 0.036 *
## TestDIMcat2030 10.4483 0.4781 477.52 < 2e-16 ***
## TestDIMcat2050 12.0588 0.4017 901.00 < 2e-16 ***
## TestDIMcat2070 11.3746 0.4376 675.70 < 2e-16 ***
## TestDIMcat2090 9.6402 0.4643 431.00 < 2e-16 ***
## TestDIMcat20110 7.7699 0.4426 308.18 < 2e-16 ***
## DOSCC 0.7748 0.1895 16.72 4.3e-05 ***
## Parity2 2.1652 0.4678 21.43 3.7e-06 ***
## Parity3 2.5068 0.5445 21.19 4.2e-06 ***
## DOMY 0.2104 0.0533 15.60 7.8e-05 ***
## TxA_SDCT:my3Mid production 0.3068 1.2690 0.06 0.809
## TxBDCT:my3Mid production -0.9791 1.0915 0.80 0.370
## TxA_SDCT:my3High production -0.4263 1.2905 0.11 0.741
## TxBDCT:my3High production -0.2135 1.1542 0.03 0.853
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 83.2 2.66
## Number of clusters: 1363 Maximum cluster size: 4out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Milk Yield"
tab## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT -0.44 (-1.96 to 1.08) Milk Yield
## 2 Low production A_SDCT - BDCT -0.55 (-2.4 to 1.3) Milk Yield
## 3 Mid production C_SDCT - BDCT 0.54 (-0.97 to 2.05) Milk Yield
## 4 Mid production A_SDCT - BDCT 0.74 (-0.83 to 2.3) Milk Yield
## 5 High production C_SDCT - BDCT -0.23 (-1.91 to 1.45) Milk Yield
## 6 High production A_SDCT - BDCT -0.76 (-2.52 to 1) Milk YieldStratifying to only include USA cows
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + DOMY,
data=dhi %>% filter(!is.na(DOSCC) & source=="US")
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## DOMY, data = dhi %>% filter(!is.na(DOSCC) & source == "US"),
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 29.422307 1.684163 305.20 < 2e-16 ***
## TxA_SDCT -1.534068 1.022797 2.25 0.13365
## TxBDCT 0.417250 0.912267 0.21 0.64740
## my3Mid production -0.687125 1.088984 0.40 0.52806
## my3High production 1.001085 1.395710 0.51 0.47321
## TestDIMcat2030 11.199074 0.512109 478.23 < 2e-16 ***
## TestDIMcat2050 12.847842 0.426208 908.69 < 2e-16 ***
## TestDIMcat2070 12.020908 0.464903 668.57 < 2e-16 ***
## TestDIMcat2090 10.476140 0.500353 438.38 < 2e-16 ***
## TestDIMcat20110 8.388605 0.475187 311.64 < 2e-16 ***
## DOSCC 0.729068 0.204831 12.67 0.00037 ***
## Parity2 1.686000 0.502318 11.27 0.00079 ***
## Parity3 2.405592 0.587740 16.75 4.3e-05 ***
## DOMY 0.228885 0.055732 16.87 4.0e-05 ***
## TxA_SDCT:my3Mid production 1.148374 1.336622 0.74 0.39025
## TxBDCT:my3Mid production -1.264223 1.247883 1.03 0.31102
## TxA_SDCT:my3High production 0.790430 1.342647 0.35 0.55606
## TxBDCT:my3High production -0.000547 1.261011 0.00 0.99965
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 84.9 2.9
## Number of clusters: 1177 Maximum cluster size: 4out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Milk Yield"
tab## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT -0.42 (-2.21 to 1.37) Milk Yield
## 2 Low production A_SDCT - BDCT -1.95 (-3.91 to 0.01) Milk Yield
## 3 Mid production C_SDCT - BDCT 0.85 (-0.82 to 2.52) Milk Yield
## 4 Mid production A_SDCT - BDCT 0.46 (-1.2 to 2.13) Milk Yield
## 5 High production C_SDCT - BDCT -0.42 (-2.13 to 1.3) Milk Yield
## 6 High production A_SDCT - BDCT -1.16 (-2.93 to 0.61) Milk Yieldfulltab <- rbind(tab.cm,
tab.cull,
tab.imipc,
tab.imicure,
tab.newimi,
tab.scc,
tab.my)
full.main <- rbind(cm.main,
cull.main,
imipc.main,
imicure.main,
iminew.main,
scc.main,
my.main)
int.p <- rbind(int.p.cm,
int.p.cull,
int.p.imipc,
int.p.cure,
int.p.newimi,
int.p.scc,
int.p.my)
write.csv(int.p,"interaction.csv")
full.main$contrast <- full.main$term
full.main$term <- NULL
tableout <- rbind(fulltab,full.main) %>% arrange(outcome,production)
write.csv(tableout,'tableout.csv')Milk yield at DO —
IMI at 1-13 DIM
Need to choose covariates
ggplot(qtr, aes(x=DOMY, y=imipc)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
IMI in early lactation
gee <- geeglm(imipc ~ my3, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = imipc ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.21483 0.03640 34.83 3.6e-09 ***
## my3Mid production -0.00362 0.01539 0.06 0.81
## my3High production 0.01883 0.02272 0.69 0.41
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.162
## Number of clusters: 16 Maximum cluster size: 1270gee <- geeglm(imipc ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = imipc ~ my3 + factor(Parity) + DOSCC + Tx, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.07067 0.04971 2.02 0.1551
## my3Mid production 0.01972 0.01479 1.78 0.1824
## my3High production 0.04586 0.02069 4.91 0.0267 *
## factor(Parity)2 0.01015 0.02261 0.20 0.6537
## factor(Parity)3 0.01602 0.01536 1.09 0.2969
## DOSCC 0.02942 0.00912 10.42 0.0012 **
## TxA_SDCT -0.00632 0.01388 0.21 0.6488
## TxBDCT -0.02268 0.02224 1.04 0.3080
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.162
## Number of clusters: 16 Maximum cluster size: 1270tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.0707 0.0497 2.02 0.155 -0.0268 0.168
## 2 my3Mid production 0.0197 0.0148 1.78 0.182 -0.00927 0.0487
## 3 my3High production 0.0459 0.0207 4.91 0.0267 0.00531 0.0864
## 4 factor(Parity)2 0.0101 0.0226 0.201 0.654 -0.0342 0.0545
## 5 factor(Parity)3 0.0160 0.0154 1.09 0.297 -0.0141 0.0461
## 6 DOSCC 0.0294 0.00912 10.4 0.00125 0.0116 0.0473
## 7 TxA_SDCT -0.00632 0.0139 0.207 0.649 -0.0335 0.0209
## 8 TxBDCT -0.0227 0.0222 1.04 0.308 -0.0663 0.0209IMI cure
ggplot(qtr, aes(x=DOMY, y=cure)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
#Univariable
gee <- geeglm(cure ~ my3, data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.87390 0.03263 717.21 <2e-16 ***
## my3Mid production 0.00785 0.02043 0.15 0.70
## my3High production 0.00962 0.02778 0.12 0.73
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.301
## Number of clusters: 16 Maximum cluster size: 432#Multivariable
gee <- geeglm(cure ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ my3 + factor(Parity) + DOSCC + Tx, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.952458 0.057031 278.92 <2e-16 ***
## my3Mid production 0.002327 0.018367 0.02 0.899
## my3High production -0.001926 0.024822 0.01 0.938
## factor(Parity)2 -0.034728 0.015091 5.30 0.021 *
## factor(Parity)3 -0.025484 0.026793 0.90 0.342
## DOSCC -0.010907 0.006862 2.53 0.112
## TxA_SDCT -0.000967 0.013113 0.01 0.941
## TxBDCT -0.010846 0.017682 0.38 0.540
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.337
## Number of clusters: 16 Maximum cluster size: 432tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.952 0.0570 279. 0 0.841 1.06
## 2 my3Mid production 0.00233 0.0184 0.0160 0.899 -0.0337 0.0383
## 3 my3High production -0.00193 0.0248 0.00602 0.938 -0.0506 0.0467
## 4 factor(Parity)2 -0.0347 0.0151 5.30 0.0214 -0.0643 -0.00515
## 5 factor(Parity)3 -0.0255 0.0268 0.905 0.342 -0.0780 0.0270
## 6 DOSCC -0.0109 0.00686 2.53 0.112 -0.0244 0.00254
## 7 TxA_SDCT -0.000967 0.0131 0.00544 0.941 -0.0267 0.0247
## 8 TxBDCT -0.0108 0.0177 0.376 0.540 -0.0455 0.0238NewIMI
ggplot(qtr, aes(x=DOMY, y=newimi)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
gee <- geeglm(newimi ~ my3, data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.21156 0.03874 29.83 4.7e-08 ***
## my3Mid production -0.01893 0.01647 1.32 0.25
## my3High production 0.00919 0.02160 0.18 0.67
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.198
## Number of clusters: 16 Maximum cluster size: 1223gee <- geeglm(newimi ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ my3 + factor(Parity) + DOSCC + Tx,
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(newimi) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.07371 0.05216 2.00 0.1576
## my3Mid production 0.00362 0.01594 0.05 0.8203
## my3High production 0.03427 0.02064 2.76 0.0969 .
## factor(Parity)2 0.01107 0.02390 0.21 0.6431
## factor(Parity)3 0.01335 0.01603 0.69 0.4049
## DOSCC 0.02839 0.00897 10.02 0.0016 **
## TxA_SDCT -0.00585 0.01637 0.13 0.7209
## TxBDCT -0.02493 0.02298 1.18 0.2779
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.198
## Number of clusters: 16 Maximum cluster size: 1223tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.0737 0.0522 2.00 0.158 -0.0285 0.176
## 2 my3Mid production 0.00362 0.0159 0.0516 0.820 -0.0276 0.0349
## 3 my3High production 0.0343 0.0206 2.76 0.0969 -0.00619 0.0747
## 4 factor(Parity)2 0.0111 0.0239 0.215 0.643 -0.0358 0.0579
## 5 factor(Parity)3 0.0134 0.0160 0.694 0.405 -0.0181 0.0448
## 6 DOSCC 0.0284 0.00897 10.0 0.00155 0.0108 0.0460
## 7 TxA_SDCT -0.00585 0.0164 0.128 0.721 -0.0379 0.0262
## 8 TxBDCT -0.0249 0.0230 1.18 0.278 -0.0700 0.0201Post calving SCC
desc <- dhi %>% group_by(my3,TestDIMcat20) %>% summarise(
atrisk = sum(!is.na(scc)),
scm.n = sum(scc > log(200)),
scm.risk = scm.n/atrisk
)
ggplot(desc, aes(x=TestDIMcat20 %>% as.character() %>% as.numeric, y=scm.risk, color = my3)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", size=1.5)+
xlab("Days in milk")+
ylab("Risk of subclinical mastitis")+
theme_bw()
mm0 <- geeglm(scc ~ my3 + TestDIMcat20 + factor(FARMID), data=dhi, id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(FARMID), data = dhi,
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 4.302244 0.132182 1059.36 < 2e-16 ***
## my3Mid production -0.107239 0.071262 2.26 0.1324
## my3High production -0.100626 0.075204 1.79 0.1809
## TestDIMcat2030 -0.517743 0.065647 62.20 3.1e-15 ***
## TestDIMcat2050 -0.559186 0.059315 88.88 < 2e-16 ***
## TestDIMcat2070 -0.410397 0.061308 44.81 2.2e-11 ***
## TestDIMcat2090 -0.279092 0.066168 17.79 2.5e-05 ***
## TestDIMcat20110 -0.122015 0.061394 3.95 0.0469 *
## factor(FARMID)2 -0.000324 0.175405 0.00 0.9985
## factor(FARMID)3 -0.064586 0.138362 0.22 0.6407
## factor(FARMID)4 0.077785 0.135469 0.33 0.5658
## factor(FARMID)5 0.132248 0.139731 0.90 0.3439
## factor(FARMID)6 -0.491592 0.181716 7.32 0.0068 **
## factor(FARMID)7 -0.347942 0.155238 5.02 0.0250 *
## factor(FARMID)8 0.024863 0.292672 0.01 0.9323
## factor(FARMID)9 -0.125526 0.296649 0.18 0.6722
## factor(FARMID)10 -0.315094 0.259063 1.48 0.2239
## factor(FARMID)11 -0.332164 0.249733 1.77 0.1835
## factor(FARMID)12 -0.600634 0.235406 6.51 0.0107 *
## factor(FARMID)13 0.101746 0.317137 0.10 0.7483
## factor(FARMID)14 0.092777 0.264044 0.12 0.7253
## factor(FARMID)15 -0.022538 0.393192 0.00 0.9543
## factor(FARMID)16 -0.166800 0.238637 0.49 0.4846
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.78 0.0595
## Number of clusters: 1364 Maximum cluster size: 4mm0 <- geeglm(scc ~ my3 + TestDIMcat20 + factor(Parity) + DOSCC + Tx + factor(FARMID), data=dhi, id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(Parity) +
## DOSCC + Tx + factor(FARMID), data = dhi, id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.332302 0.180892 339.35 < 2e-16 ***
## my3Mid production 0.042088 0.072003 0.34 0.5589
## my3High production 0.075529 0.076253 0.98 0.3219
## TestDIMcat2030 -0.541606 0.064565 70.37 < 2e-16 ***
## TestDIMcat2050 -0.556534 0.058934 89.18 < 2e-16 ***
## TestDIMcat2070 -0.418189 0.060615 47.60 5.2e-12 ***
## TestDIMcat2090 -0.286162 0.064782 19.51 1.0e-05 ***
## TestDIMcat20110 -0.124401 0.061089 4.15 0.0417 *
## factor(Parity)2 0.133325 0.062498 4.55 0.0329 *
## factor(Parity)3 0.302242 0.073518 16.90 3.9e-05 ***
## DOSCC 0.175860 0.026812 43.02 5.4e-11 ***
## TxA_SDCT -0.000571 0.068141 0.00 0.9933
## TxBDCT -0.027835 0.063889 0.19 0.6631
## factor(FARMID)2 0.054528 0.170947 0.10 0.7497
## factor(FARMID)3 -0.042897 0.134559 0.10 0.7499
## factor(FARMID)4 0.033644 0.131551 0.07 0.7981
## factor(FARMID)5 0.031388 0.135680 0.05 0.8171
## factor(FARMID)6 -0.507588 0.180322 7.92 0.0049 **
## factor(FARMID)7 -0.300935 0.149200 4.07 0.0437 *
## factor(FARMID)8 -0.067311 0.279497 0.06 0.8097
## factor(FARMID)9 -0.052484 0.298032 0.03 0.8602
## factor(FARMID)10 -0.264201 0.226903 1.36 0.2443
## factor(FARMID)11 -0.406134 0.240999 2.84 0.0919 .
## factor(FARMID)12 -0.463233 0.217017 4.56 0.0328 *
## factor(FARMID)13 0.128225 0.303021 0.18 0.6722
## factor(FARMID)14 0.098630 0.254484 0.15 0.6983
## factor(FARMID)15 0.022149 0.368868 0.00 0.9521
## factor(FARMID)16 -0.061304 0.232800 0.07 0.7923
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0566
## Number of clusters: 1364 Maximum cluster size: 4summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(Parity) +
## DOSCC + Tx + factor(FARMID), data = dhi, id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.332302 0.180892 339.35 < 2e-16 ***
## my3Mid production 0.042088 0.072003 0.34 0.5589
## my3High production 0.075529 0.076253 0.98 0.3219
## TestDIMcat2030 -0.541606 0.064565 70.37 < 2e-16 ***
## TestDIMcat2050 -0.556534 0.058934 89.18 < 2e-16 ***
## TestDIMcat2070 -0.418189 0.060615 47.60 5.2e-12 ***
## TestDIMcat2090 -0.286162 0.064782 19.51 1.0e-05 ***
## TestDIMcat20110 -0.124401 0.061089 4.15 0.0417 *
## factor(Parity)2 0.133325 0.062498 4.55 0.0329 *
## factor(Parity)3 0.302242 0.073518 16.90 3.9e-05 ***
## DOSCC 0.175860 0.026812 43.02 5.4e-11 ***
## TxA_SDCT -0.000571 0.068141 0.00 0.9933
## TxBDCT -0.027835 0.063889 0.19 0.6631
## factor(FARMID)2 0.054528 0.170947 0.10 0.7497
## factor(FARMID)3 -0.042897 0.134559 0.10 0.7499
## factor(FARMID)4 0.033644 0.131551 0.07 0.7981
## factor(FARMID)5 0.031388 0.135680 0.05 0.8171
## factor(FARMID)6 -0.507588 0.180322 7.92 0.0049 **
## factor(FARMID)7 -0.300935 0.149200 4.07 0.0437 *
## factor(FARMID)8 -0.067311 0.279497 0.06 0.8097
## factor(FARMID)9 -0.052484 0.298032 0.03 0.8602
## factor(FARMID)10 -0.264201 0.226903 1.36 0.2443
## factor(FARMID)11 -0.406134 0.240999 2.84 0.0919 .
## factor(FARMID)12 -0.463233 0.217017 4.56 0.0328 *
## factor(FARMID)13 0.128225 0.303021 0.18 0.6722
## factor(FARMID)14 0.098630 0.254484 0.15 0.6983
## factor(FARMID)15 0.022149 0.368868 0.00 0.9521
## factor(FARMID)16 -0.061304 0.232800 0.07 0.7923
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0566
## Number of clusters: 1364 Maximum cluster size: 4atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~ my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(my3,TestDIMcat20,SCC,LCL,UCL))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=my3, colour=my3) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$LCL, ymax=emm$UCL,colour=my3,fill=my3), linetype=0, alpha=0.1) + geom_line(aes(colour=my3,linetype=my3),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve
Repeat analysis with different definition for my3 (i.e within herd)
Compare agreement between my3 types
cohen.kappa(x=cbind(cow$my3,cow$my3h))## Call: cohen.kappa1(x = x, w = w, n.obs = n.obs, alpha = alpha, levels = levels)
##
## Cohen Kappa and Weighted Kappa correlation coefficients and confidence boundaries
## lower estimate upper
## unweighted kappa 0.46 0.50 0.53
## weighted kappa 0.72 0.74 0.76
##
## Number of subjects = 1485cow %>% tabyl(my3,my3h)## my3 Low production Mid production High production
## Low production 369 121 5
## Mid production 134 248 113
## High production 0 125 370Create new my3 variable
qtr$my3 <- qtr$my3h
cow$my3 <- cow$my3h
dhi$my3 <- dhi$my3htable1(~ Tx + factor(imido) + DOMY + DOSCC + factor(Parity) + factor(FARMID) + source | my3, data=qtr, na.is.category = F)| Low production (N=1705) |
Mid production (N=1698) |
High production (N=1694) |
Overall (N=5097) |
|
|---|---|---|---|---|
| Tx | ||||
| C_SDCT | 561 (32.9%) | 610 (35.9%) | 678 (40.0%) | 1849 (36.3%) |
| A_SDCT | 439 (25.7%) | 499 (29.4%) | 417 (24.6%) | 1355 (26.6%) |
| BDCT | 705 (41.3%) | 589 (34.7%) | 599 (35.4%) | 1893 (37.1%) |
| factor(imido) | ||||
| 0 | 1169 (75.1%) | 1194 (75.3%) | 1138 (73.6%) | 3501 (74.7%) |
| 1 | 387 (24.9%) | 392 (24.7%) | 408 (26.4%) | 1187 (25.3%) |
| Missing | 149 (8.7%) | 112 (6.6%) | 148 (8.7%) | 409 (8.0%) |
| DOMY | ||||
| Mean (SD) | 18.4 (6.60) | 27.0 (4.89) | 34.2 (5.20) | 26.5 (8.57) |
| Median [Min, Max] | 19.5 [1.81, 29.0] | 28.6 [13.6, 34.0] | 34.5 [21.6, 49.4] | 27.2 [1.81, 49.4] |
| DOSCC | ||||
| Mean (SD) | 4.93 (1.24) | 4.29 (1.11) | 4.05 (1.09) | 4.42 (1.20) |
| Median [Min, Max] | 4.87 [1.61, 8.35] | 4.19 [1.61, 7.69] | 3.91 [1.61, 8.59] | 4.33 [1.61, 8.59] |
| Missing | 0 (0%) | 0 (0%) | 4 (0.2%) | 4 (0.1%) |
| factor(Parity) | ||||
| 1 | 574 (33.7%) | 793 (46.7%) | 869 (51.3%) | 2236 (43.9%) |
| 2 | 580 (34.0%) | 445 (26.2%) | 447 (26.4%) | 1472 (28.9%) |
| 3 | 551 (32.3%) | 460 (27.1%) | 378 (22.3%) | 1389 (27.3%) |
| factor(FARMID) | ||||
| 1 | 99 (5.81%) | 98 (5.77%) | 98 (5.79%) | 295 (5.79%) |
| 2 | 108 (6.33%) | 108 (6.36%) | 108 (6.38%) | 324 (6.36%) |
| 3 | 279 (16.4%) | 278 (16.4%) | 278 (16.4%) | 835 (16.4%) |
| 4 | 424 (24.9%) | 423 (24.9%) | 423 (25.0%) | 1270 (24.9%) |
| 5 | 296 (17.4%) | 296 (17.4%) | 295 (17.4%) | 887 (17.4%) |
| 6 | 56 (3.28%) | 56 (3.30%) | 56 (3.31%) | 168 (3.30%) |
| 7 | 132 (7.74%) | 131 (7.71%) | 131 (7.73%) | 394 (7.73%) |
| 8 | 23 (1.35%) | 23 (1.35%) | 22 (1.30%) | 68 (1.33%) |
| 9 | 26 (1.52%) | 25 (1.47%) | 25 (1.48%) | 76 (1.49%) |
| 10 | 27 (1.58%) | 27 (1.59%) | 27 (1.59%) | 81 (1.59%) |
| 11 | 26 (1.52%) | 26 (1.53%) | 26 (1.53%) | 78 (1.53%) |
| 12 | 23 (1.35%) | 23 (1.35%) | 22 (1.30%) | 68 (1.33%) |
| 13 | 32 (1.88%) | 31 (1.83%) | 31 (1.83%) | 94 (1.84%) |
| 14 | 83 (4.87%) | 83 (4.89%) | 82 (4.84%) | 248 (4.87%) |
| 15 | 19 (1.11%) | 18 (1.06%) | 18 (1.06%) | 55 (1.08%) |
| 16 | 52 (3.05%) | 52 (3.06%) | 52 (3.07%) | 156 (3.06%) |
| source | ||||
| CA | 311 (18.2%) | 308 (18.1%) | 305 (18.0%) | 924 (18.1%) |
| US | 1394 (81.8%) | 1390 (81.9%) | 1389 (82.0%) | 4173 (81.9%) |
table1(~ factor(imido) + DOMY + DOSCC + factor(Parity) | my3*Tx, data=qtr, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=561) |
A_SDCT (N=439) |
BDCT (N=705) |
C_SDCT (N=610) |
A_SDCT (N=499) |
BDCT (N=589) |
C_SDCT (N=678) |
A_SDCT (N=417) |
BDCT (N=599) |
C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
|
| factor(imido) | ||||||||||||
| 0 | 392 (75.4%) | 301 (77.2%) | 476 (73.7%) | 411 (73.1%) | 361 (76.0%) | 422 (76.9%) | 471 (75.2%) | 276 (73.4%) | 391 (71.9%) | 1274 (74.6%) | 938 (75.6%) | 1289 (74.1%) |
| 1 | 128 (24.6%) | 89 (22.8%) | 170 (26.3%) | 151 (26.9%) | 114 (24.0%) | 127 (23.1%) | 155 (24.8%) | 100 (26.6%) | 153 (28.1%) | 434 (25.4%) | 303 (24.4%) | 450 (25.9%) |
| Missing | 41 (7.3%) | 49 (11.2%) | 59 (8.4%) | 48 (7.9%) | 24 (4.8%) | 40 (6.8%) | 52 (7.7%) | 41 (9.8%) | 55 (9.2%) | 141 (7.6%) | 114 (8.4%) | 154 (8.1%) |
| DOMY | ||||||||||||
| Mean (SD) | 18.3 (6.54) | 18.8 (7.04) | 18.2 (6.35) | 26.6 (5.19) | 27.8 (4.67) | 26.6 (4.66) | 34.0 (5.36) | 35.0 (4.36) | 34.0 (5.50) | 26.8 (8.56) | 27.1 (8.45) | 25.8 (8.61) |
| Median [Min, Max] | 19.1 [1.81, 27.7] | 21.3 [2.72, 29.0] | 18.6 [4.54, 28.1] | 28.6 [13.6, 34.0] | 29.5 [13.6, 34.0] | 27.6 [14.5, 34.0] | 34.0 [21.6, 49.4] | 35.4 [26.3, 49.4] | 34.0 [22.4, 49.4] | 27.7 [1.81, 49.4] | 28.6 [2.72, 49.4] | 26.4 [4.54, 49.4] |
| DOSCC | ||||||||||||
| Mean (SD) | 4.98 (1.29) | 4.82 (1.21) | 4.96 (1.21) | 4.21 (1.08) | 4.38 (1.16) | 4.29 (1.09) | 4.00 (1.10) | 4.09 (1.11) | 4.09 (1.05) | 4.36 (1.22) | 4.43 (1.20) | 4.48 (1.18) |
| Median [Min, Max] | 4.95 [1.95, 8.17] | 4.81 [1.61, 7.38] | 4.87 [2.48, 8.35] | 4.13 [1.61, 7.69] | 4.30 [2.20, 7.56] | 4.26 [1.61, 7.16] | 3.89 [1.79, 8.59] | 3.97 [2.30, 7.59] | 4.06 [1.61, 7.38] | 4.20 [1.61, 8.59] | 4.38 [1.61, 7.59] | 4.40 [1.61, 8.35] |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 4 (0.7%) | 0 (0%) | 0 (0%) | 4 (0.2%) |
| factor(Parity) | ||||||||||||
| 1 | 185 (33.0%) | 128 (29.2%) | 261 (37.0%) | 323 (53.0%) | 234 (46.9%) | 236 (40.1%) | 364 (53.7%) | 237 (56.8%) | 268 (44.7%) | 872 (47.2%) | 599 (44.2%) | 765 (40.4%) |
| 2 | 190 (33.9%) | 174 (39.6%) | 216 (30.6%) | 125 (20.5%) | 146 (29.3%) | 174 (29.5%) | 157 (23.2%) | 108 (25.9%) | 182 (30.4%) | 472 (25.5%) | 428 (31.6%) | 572 (30.2%) |
| 3 | 186 (33.2%) | 137 (31.2%) | 228 (32.3%) | 162 (26.6%) | 119 (23.8%) | 179 (30.4%) | 157 (23.2%) | 72 (17.3%) | 149 (24.9%) | 505 (27.3%) | 328 (24.2%) | 556 (29.4%) |
table1(~ factor(imipc) + factor(cure) + factor(newimi) | my3*Tx, data=qtr, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=561) |
A_SDCT (N=439) |
BDCT (N=705) |
C_SDCT (N=610) |
A_SDCT (N=499) |
BDCT (N=589) |
C_SDCT (N=678) |
A_SDCT (N=417) |
BDCT (N=599) |
C_SDCT (N=1849) |
A_SDCT (N=1355) |
BDCT (N=1893) |
|
| factor(imipc) | ||||||||||||
| 0 | 426 (75.9%) | 342 (77.9%) | 553 (78.4%) | 470 (77.0%) | 389 (78.0%) | 472 (80.1%) | 529 (78.0%) | 321 (77.0%) | 475 (79.3%) | 1425 (77.1%) | 1052 (77.6%) | 1500 (79.2%) |
| 1 | 135 (24.1%) | 97 (22.1%) | 152 (21.6%) | 140 (23.0%) | 110 (22.0%) | 117 (19.9%) | 149 (22.0%) | 96 (23.0%) | 124 (20.7%) | 424 (22.9%) | 303 (22.4%) | 393 (20.8%) |
| factor(cure) | ||||||||||||
| 0 | 15 (11.7%) | 12 (13.5%) | 24 (14.1%) | 19 (12.6%) | 10 (8.77%) | 12 (9.45%) | 15 (9.68%) | 14 (14.0%) | 21 (13.7%) | 49 (11.3%) | 36 (11.9%) | 57 (12.7%) |
| 1 | 113 (88.3%) | 77 (86.5%) | 146 (85.9%) | 132 (87.4%) | 104 (91.2%) | 115 (90.6%) | 140 (90.3%) | 86 (86.0%) | 132 (86.3%) | 385 (88.7%) | 267 (88.1%) | 393 (87.3%) |
| Missing | 433 (77.2%) | 350 (79.7%) | 535 (75.9%) | 459 (75.2%) | 385 (77.2%) | 462 (78.4%) | 523 (77.1%) | 317 (76.0%) | 446 (74.5%) | 1415 (76.5%) | 1052 (77.6%) | 1443 (76.2%) |
| factor(newimi) | ||||||||||||
| 0 | 399 (77.0%) | 302 (78.2%) | 510 (79.6%) | 433 (78.3%) | 373 (79.2%) | 440 (80.7%) | 491 (78.9%) | 289 (78.7%) | 436 (81.5%) | 1323 (78.1%) | 964 (78.8%) | 1386 (80.5%) |
| 1 | 119 (23.0%) | 84 (21.8%) | 131 (20.4%) | 120 (21.7%) | 98 (20.8%) | 105 (19.3%) | 131 (21.1%) | 78 (21.3%) | 99 (18.5%) | 370 (21.9%) | 260 (21.2%) | 335 (19.5%) |
| Missing | 43 (7.7%) | 53 (12.1%) | 64 (9.1%) | 57 (9.3%) | 28 (5.6%) | 44 (7.5%) | 56 (8.3%) | 50 (12.0%) | 64 (10.7%) | 156 (8.4%) | 131 (9.7%) | 172 (9.1%) |
table1(~ factor(CM) + factor(CMDP) | my3*Tx, data=cow, na.is.category = F)Low production |
Mid production |
High production |
Overall |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C_SDCT (N=169) |
A_SDCT (N=128) |
BDCT (N=206) |
C_SDCT (N=172) |
A_SDCT (N=147) |
BDCT (N=175) |
C_SDCT (N=195) |
A_SDCT (N=119) |
BDCT (N=174) |
C_SDCT (N=536) |
A_SDCT (N=394) |
BDCT (N=555) |
|
| factor(CM) | ||||||||||||
| 0 | 153 (90.5%) | 109 (85.2%) | 173 (84.0%) | 149 (86.6%) | 132 (89.8%) | 148 (84.6%) | 170 (87.2%) | 105 (88.2%) | 143 (82.2%) | 472 (88.1%) | 346 (87.8%) | 464 (83.6%) |
| 1 | 16 (9.47%) | 19 (14.8%) | 33 (16.0%) | 23 (13.4%) | 15 (10.2%) | 27 (15.4%) | 25 (12.8%) | 14 (11.8%) | 31 (17.8%) | 64 (11.9%) | 48 (12.2%) | 91 (16.4%) |
| factor(CMDP) | ||||||||||||
| 0 | 169 (100%) | 128 (100%) | 206 (100%) | 171 (99.4%) | 147 (100%) | 175 (100%) | 194 (99.5%) | 116 (97.5%) | 173 (99.4%) | 534 (99.6%) | 391 (99.2%) | 554 (99.8%) |
| 1 | 0 (0%) | 0 (0%) | 0 (0%) | 1 (0.581%) | 0 (0%) | 0 (0%) | 1 (0.513%) | 3 (2.52%) | 1 (0.575%) | 2 (0.373%) | 3 (0.761%) | 1 (0.180%) |
cow %>% tabyl(source,my3) %>% adorn_totals("row") %>%
adorn_percentages("row") %>%
adorn_pct_formatting(digits = 1) %>%
adorn_ns %>%
adorn_title## my3
## source Low production Mid production High production
## CA 35.0% (96) 32.8% (90) 32.1% (88)
## US 33.6% (407) 33.4% (404) 33.0% (400)
## Total 33.9% (503) 33.3% (494) 32.9% (488)cow %>% tabyl(FARMID,my3) %>% adorn_totals("row") %>%
adorn_percentages("row") %>%
adorn_pct_formatting(digits = 1) %>%
adorn_ns %>%
adorn_title## my3
## FARMID Low production Mid production High production
## 1 34.2% (26) 32.9% (25) 32.9% (25)
## 2 34.1% (28) 32.9% (27) 32.9% (27)
## 3 33.5% (88) 33.5% (88) 33.1% (87)
## 4 33.5% (130) 33.2% (129) 33.2% (129)
## 5 33.5% (81) 33.5% (81) 33.1% (80)
## 6 34.0% (16) 34.0% (16) 31.9% (15)
## 7 33.6% (38) 33.6% (38) 32.7% (37)
## 8 33.3% (6) 33.3% (6) 33.3% (6)
## 9 36.0% (9) 32.0% (8) 32.0% (8)
## 10 34.8% (8) 34.8% (8) 30.4% (7)
## 11 36.4% (8) 31.8% (7) 31.8% (7)
## 12 36.8% (7) 31.6% (6) 31.6% (6)
## 13 34.5% (10) 34.5% (10) 31.0% (9)
## 14 34.2% (26) 32.9% (25) 32.9% (25)
## 15 37.5% (6) 31.2% (5) 31.2% (5)
## 16 34.8% (16) 32.6% (15) 32.6% (15)
## Total 33.9% (503) 33.3% (494) 32.9% (488)Extract crude frequencies
cm.freq.tx <- cow %>% group_by(Tx) %>% summarise(
n.cm.0 = sum(CM==0),
n.cm.1 = sum(CM==1),
p.cm.1 = n.cm.1/(n.cm.1+n.cm.0),
cm = paste0(n.cm.1, "/" ,n.cm.0 + n.cm.1, " (",100*round(p.cm.1,4),"%)")
)
cm.freq.tx$my3 <- "All production levels"
cm.freq.my3 <- cow %>% group_by(my3,Tx) %>% summarise(
n.cm.0 = sum(CM==0),
n.cm.1 = sum(CM==1),
p.cm.1 = n.cm.1/(n.cm.1+n.cm.0),
cm = paste0(n.cm.1, "/" ,n.cm.0 + n.cm.1, " (",100*round(p.cm.1,4),"%)")
)
cm.freq <- rbind(cm.freq.tx,cm.freq.my3) %>% select(my3,Tx,cm)
cm.freq$risk <- cm.freq$cm
cm.freq$outcome <- "Clinical mastitis"
cm.freq$cm <- NULL
#Cull
Cull2.freq.tx <- cow %>% group_by(Tx) %>% summarise(
n.Cull2.0 = sum(Cull2==0),
n.Cull2.1 = sum(Cull2==1),
p.Cull2.1 = n.Cull2.1/(n.Cull2.1+n.Cull2.0),
Cull2 = paste0(n.Cull2.1, "/" ,n.Cull2.0 + n.Cull2.1, " (",100*round(p.Cull2.1,4),"%)")
)
Cull2.freq.tx$my3 <- "All production levels"
Cull2.freq.my3 <- cow %>% group_by(my3,Tx) %>% summarise(
n.Cull2.0 = sum(Cull2==0),
n.Cull2.1 = sum(Cull2==1),
p.Cull2.1 = n.Cull2.1/(n.Cull2.1+n.Cull2.0),
Cull2 = paste0(n.Cull2.1, "/" ,n.Cull2.0 + n.Cull2.1, " (",100*round(p.Cull2.1,4),"%)")
)
Cull2.freq <- rbind(Cull2.freq.tx,Cull2.freq.my3) %>% select(my3,Tx,Cull2)
Cull2.freq$risk <- Cull2.freq$Cull2
Cull2.freq$outcome <- "Cull"
Cull2.freq$Cull2 <- NULL
imi.freq.tx <- qtr %>% group_by(Tx) %>% summarise(
n.imipc.0 = sum(imipc==0),
n.imipc.1 = sum(imipc==1),
p.imipc.1 = n.imipc.1/(n.imipc.1+n.imipc.0),
imipc = paste0(n.imipc.1, "/" ,n.imipc.0 + n.imipc.1, " (",100*round(p.imipc.1,4),"%)"),
n.newimi.0 = sum(newimi==0,na.rm=T),
n.newimi.1 = sum(newimi==1,na.rm=T),
p.newimi.1 = n.newimi.1/(n.newimi.1+n.newimi.0),
newimi = paste0(n.newimi.1, "/" ,n.newimi.0 + n.newimi.1, " (",100*round(p.newimi.1,4),"%)"),
n.cure.0 = sum(cure==0,na.rm=T),
n.cure.1 = sum(cure==1,na.rm=T),
p.cure.1 = n.cure.1/(n.cure.1+n.cure.0),
imicure = paste0(n.cure.1, "/" ,n.cure.0 + n.cure.1, " (",100*round(p.cure.1,4),"%)"),
)
imi.freq.tx$my3 <- "All production levels"
imi.freq.my3 <- qtr %>% group_by(my3,Tx) %>% summarise(
n.imipc.0 = sum(imipc==0),
n.imipc.1 = sum(imipc==1),
p.imipc.1 = n.imipc.1/(n.imipc.1+n.imipc.0),
imipc = paste0(n.imipc.1, "/" ,n.imipc.0 + n.imipc.1, " (",100*round(p.imipc.1,4),"%)"),
n.newimi.0 = sum(newimi==0,na.rm=T),
n.newimi.1 = sum(newimi==1,na.rm=T),
p.newimi.1 = n.newimi.1/(n.newimi.1+n.newimi.0),
newimi = paste0(n.newimi.1, "/" ,n.newimi.0 + n.newimi.1, " (",100*round(p.newimi.1,4),"%)"),
n.cure.0 = sum(cure==0,na.rm=T),
n.cure.1 = sum(cure==1,na.rm=T),
p.cure.1 = n.cure.1/(n.cure.1+n.cure.0),
imicure = paste0(n.cure.1, "/" ,n.cure.0 + n.cure.1, " (",100*round(p.cure.1,4),"%)"),
)
colnames(imi.freq.tx)## [1] "Tx" "n.imipc.0" "n.imipc.1" "p.imipc.1" "imipc"
## [6] "n.newimi.0" "n.newimi.1" "p.newimi.1" "newimi" "n.cure.0"
## [11] "n.cure.1" "p.cure.1" "imicure" "my3"colnames(imi.freq.my3)## [1] "my3" "Tx" "n.imipc.0" "n.imipc.1" "p.imipc.1"
## [6] "imipc" "n.newimi.0" "n.newimi.1" "p.newimi.1" "newimi"
## [11] "n.cure.0" "n.cure.1" "p.cure.1" "imicure"imi.freq <- rbind(imi.freq.tx,imi.freq.my3) %>% select(my3,Tx,imipc,newimi,imicure) %>% pivot_longer(
cols = imipc:imicure,
names_to = "outcome",
values_to = "risk"
)
freq.tab <- rbind(imi.freq,cm.freq,Cull2.freq) %>% arrange(outcome,my3)
write.csv(freq.tab,"freq.table2.csv")Analysis
Post calving clinical mastitis
library(survival)
library(survminer)
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow)
summary(KM)## Call: survfit(formula = Surv(CMTAR, CM) ~ Tx, data = cow)
##
## Tx=C_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 2 535 2 0.996 0.00264 0.991 1.000
## 3 531 1 0.994 0.00323 0.988 1.000
## 4 528 2 0.991 0.00418 0.982 0.999
## 9 522 1 0.989 0.00458 0.980 0.998
## 11 520 1 0.987 0.00495 0.977 0.997
## 15 517 1 0.985 0.00529 0.975 0.995
## 16 516 1 0.983 0.00562 0.972 0.994
## 18 515 1 0.981 0.00592 0.970 0.993
## 21 514 1 0.979 0.00621 0.967 0.991
## 24 511 1 0.977 0.00649 0.965 0.990
## 25 509 1 0.975 0.00675 0.962 0.989
## 27 508 1 0.973 0.00701 0.960 0.987
## 31 507 1 0.972 0.00725 0.957 0.986
## 32 506 1 0.970 0.00749 0.955 0.984
## 33 505 1 0.968 0.00771 0.953 0.983
## 35 502 1 0.966 0.00794 0.950 0.981
## 38 500 1 0.964 0.00815 0.948 0.980
## 40 498 2 0.960 0.00857 0.943 0.977
## 42 496 1 0.958 0.00877 0.941 0.975
## 43 494 1 0.956 0.00896 0.939 0.974
## 47 491 1 0.954 0.00915 0.936 0.972
## 50 489 1 0.952 0.00934 0.934 0.971
## 52 488 1 0.950 0.00952 0.932 0.969
## 53 487 1 0.948 0.00970 0.929 0.967
## 58 485 1 0.946 0.00987 0.927 0.966
## 59 483 1 0.944 0.01005 0.925 0.964
## 60 482 1 0.942 0.01021 0.923 0.963
## 66 480 1 0.940 0.01038 0.920 0.961
## 70 479 1 0.938 0.01054 0.918 0.959
## 71 477 3 0.933 0.01101 0.911 0.954
## 72 474 1 0.931 0.01116 0.909 0.953
## 73 473 1 0.929 0.01131 0.907 0.951
## 76 470 1 0.927 0.01146 0.904 0.949
## 77 468 1 0.925 0.01160 0.902 0.948
## 78 467 1 0.923 0.01175 0.900 0.946
## 80 465 1 0.921 0.01189 0.898 0.944
## 83 464 1 0.919 0.01203 0.895 0.943
## 84 463 1 0.917 0.01216 0.893 0.941
## 86 461 1 0.915 0.01230 0.891 0.939
## 88 460 1 0.913 0.01243 0.889 0.937
## 94 459 1 0.911 0.01256 0.886 0.936
## 95 458 1 0.909 0.01269 0.884 0.934
## 97 456 2 0.905 0.01295 0.880 0.931
## 100 454 2 0.901 0.01319 0.875 0.927
## 101 452 1 0.899 0.01331 0.873 0.925
## 103 451 1 0.897 0.01343 0.871 0.924
## 105 450 1 0.895 0.01355 0.869 0.922
## 106 449 1 0.893 0.01366 0.866 0.920
## 108 448 1 0.891 0.01378 0.864 0.918
## 110 447 1 0.889 0.01389 0.862 0.917
## 112 445 1 0.887 0.01400 0.860 0.915
## 115 444 2 0.883 0.01422 0.855 0.911
## 117 439 3 0.877 0.01454 0.849 0.906
## 120 434 1 0.875 0.01465 0.847 0.904
##
## Tx=A_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 1 394 2 0.995 0.00358 0.988 1.000
## 2 390 1 0.992 0.00439 0.984 1.000
## 3 388 2 0.987 0.00566 0.976 0.998
## 6 384 1 0.985 0.00620 0.973 0.997
## 15 379 2 0.979 0.00718 0.966 0.994
## 19 377 1 0.977 0.00761 0.962 0.992
## 24 374 2 0.972 0.00842 0.955 0.988
## 29 371 1 0.969 0.00880 0.952 0.986
## 31 370 1 0.966 0.00915 0.949 0.985
## 35 368 2 0.961 0.00983 0.942 0.981
## 37 366 1 0.959 0.01015 0.939 0.979
## 42 364 1 0.956 0.01046 0.936 0.977
## 43 362 2 0.951 0.01104 0.929 0.973
## 45 359 1 0.948 0.01133 0.926 0.970
## 46 358 2 0.943 0.01187 0.920 0.966
## 50 356 1 0.940 0.01212 0.917 0.964
## 51 355 2 0.935 0.01262 0.910 0.960
## 53 351 2 0.929 0.01310 0.904 0.955
## 54 349 1 0.927 0.01333 0.901 0.953
## 56 346 1 0.924 0.01356 0.898 0.951
## 57 345 1 0.921 0.01378 0.895 0.949
## 58 344 1 0.919 0.01400 0.892 0.947
## 60 342 1 0.916 0.01421 0.889 0.944
## 61 341 1 0.913 0.01442 0.886 0.942
## 62 340 1 0.911 0.01463 0.882 0.940
## 71 339 2 0.905 0.01503 0.876 0.935
## 75 337 2 0.900 0.01541 0.870 0.931
## 76 335 2 0.895 0.01578 0.864 0.926
## 81 333 1 0.892 0.01596 0.861 0.924
## 90 329 1 0.889 0.01614 0.858 0.921
## 99 327 2 0.884 0.01649 0.852 0.917
## 107 323 1 0.881 0.01667 0.849 0.914
## 111 322 1 0.878 0.01684 0.846 0.912
## 112 321 1 0.876 0.01701 0.843 0.909
## 119 319 1 0.873 0.01717 0.840 0.907
##
## Tx=BDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 0 555 1 0.998 0.00180 0.995 1.000
## 1 554 2 0.995 0.00311 0.989 1.000
## 2 552 1 0.993 0.00359 0.986 1.000
## 3 551 3 0.987 0.00474 0.978 0.997
## 4 547 1 0.986 0.00506 0.976 0.996
## 5 545 2 0.982 0.00565 0.971 0.993
## 6 542 1 0.980 0.00592 0.969 0.992
## 7 539 1 0.978 0.00619 0.966 0.991
## 9 536 1 0.977 0.00644 0.964 0.989
## 10 534 1 0.975 0.00668 0.962 0.988
## 11 533 2 0.971 0.00714 0.957 0.985
## 14 529 1 0.969 0.00736 0.955 0.984
## 16 528 1 0.967 0.00757 0.953 0.982
## 17 527 1 0.966 0.00777 0.950 0.981
## 18 526 1 0.964 0.00797 0.948 0.979
## 19 525 2 0.960 0.00835 0.944 0.977
## 20 523 3 0.955 0.00889 0.937 0.972
## 22 518 2 0.951 0.00923 0.933 0.969
## 26 514 1 0.949 0.00940 0.931 0.968
## 27 513 1 0.947 0.00956 0.929 0.966
## 31 509 1 0.945 0.00972 0.926 0.965
## 33 508 1 0.943 0.00988 0.924 0.963
## 34 507 1 0.942 0.01003 0.922 0.961
## 35 506 1 0.940 0.01018 0.920 0.960
## 36 505 1 0.938 0.01033 0.918 0.958
## 37 503 1 0.936 0.01048 0.916 0.957
## 38 502 2 0.932 0.01076 0.911 0.954
## 40 499 1 0.930 0.01090 0.909 0.952
## 42 496 1 0.928 0.01104 0.907 0.950
## 47 495 1 0.927 0.01118 0.905 0.949
## 50 493 1 0.925 0.01131 0.903 0.947
## 52 492 1 0.923 0.01144 0.901 0.946
## 53 491 2 0.919 0.01170 0.896 0.942
## 57 489 1 0.917 0.01183 0.894 0.941
## 58 488 2 0.913 0.01207 0.890 0.937
## 59 486 1 0.912 0.01219 0.888 0.936
## 61 485 1 0.910 0.01231 0.886 0.934
## 63 483 1 0.908 0.01243 0.884 0.932
## 64 482 1 0.906 0.01255 0.882 0.931
## 66 481 1 0.904 0.01266 0.880 0.929
## 68 480 2 0.900 0.01288 0.875 0.926
## 70 478 1 0.898 0.01299 0.873 0.924
## 73 476 1 0.896 0.01310 0.871 0.923
## 74 475 2 0.893 0.01332 0.867 0.919
## 76 473 1 0.891 0.01342 0.865 0.918
## 78 472 2 0.887 0.01363 0.861 0.914
## 79 469 1 0.885 0.01373 0.859 0.912
## 81 468 1 0.883 0.01383 0.857 0.911
## 82 467 1 0.881 0.01393 0.855 0.909
## 83 466 1 0.879 0.01403 0.852 0.907
## 84 465 2 0.876 0.01422 0.848 0.904
## 87 463 1 0.874 0.01431 0.846 0.902
## 91 462 2 0.870 0.01450 0.842 0.899
## 92 460 2 0.866 0.01468 0.838 0.896
## 95 457 2 0.862 0.01486 0.834 0.892
## 96 455 1 0.861 0.01495 0.832 0.890
## 97 454 1 0.859 0.01503 0.830 0.889
## 101 453 1 0.857 0.01512 0.828 0.887
## 102 452 1 0.855 0.01521 0.826 0.885
## 103 451 2 0.851 0.01537 0.821 0.882
## 104 449 1 0.849 0.01545 0.819 0.880
## 107 448 1 0.847 0.01554 0.817 0.878
## 109 447 1 0.845 0.01562 0.815 0.877
## 110 446 1 0.844 0.01570 0.813 0.875
## 111 445 1 0.842 0.01577 0.811 0.873
## 112 444 1 0.840 0.01585 0.809 0.871
## 113 443 2 0.836 0.01601 0.805 0.868
## 114 441 1 0.834 0.01608 0.803 0.866
## 115 440 2 0.830 0.01623 0.799 0.863ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
Clinical mastitis - all production groups combined
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow)
km.all <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Low production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="Low production"))
km.low <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "Low production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Mid production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="Mid production"))
km.mid <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "Mid production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)High production cows
KM <- survfit(Surv(CMTAR, CM) ~ Tx, data = cow %>% filter(my3=="High production"))
km.high <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.20),
legend = "bottom",
title = "High production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)#png("plot.png", width=500*10, height=720*10, res=100*15)
splots <- list(km.all,
km.low,
km.mid,
km.high)
arrange_ggsurvplots(splots,ncol=2,nrow=2)
#dev.off()Cox regression
All production groups pooled together
library(broom)
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
SR <- coxph(Surv(CMTAR, CM) ~ Tx + my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(SR)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(CMTAR, CM)
## Df Chisq Pr(>Chisq)
## Tx 2 4.72 0.094 .
## my3 2 2.60 0.273
## factor(Parity) 2 8.42 0.015 *
## DOSCC 1 1.27 0.259
## factor(source) 1 0.83 0.363
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(SR)## Call:
## coxph(formula = Surv(CMTAR, CM) ~ Tx + my3 + factor(Parity) +
## DOSCC + factor(source), data = cow, cluster = FARMID)
##
## n= 1484, number of events= 203
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z Pr(>|z|)
## TxC_SDCT -0.3250 0.7225 0.1636 0.1667 -1.95 0.0512 .
## TxA_SDCT -0.2295 0.7950 0.1868 0.1452 -1.58 0.1140
## my3Mid production 0.0664 1.0687 0.1777 0.2351 0.28 0.7776
## my3High production 0.2005 1.2221 0.1783 0.2221 0.90 0.3666
## factor(Parity)2 0.3213 1.3790 0.1745 0.1110 2.89 0.0038 **
## factor(Parity)3 0.2936 1.3412 0.1795 0.2302 1.28 0.2021
## DOSCC 0.1171 1.1242 0.0627 0.1037 1.13 0.2591
## factor(source)US -0.2463 0.7817 0.1763 0.2708 -0.91 0.3631
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 0.723 1.384 0.521 1.00
## TxA_SDCT 0.795 1.258 0.598 1.06
## my3Mid production 1.069 0.936 0.674 1.69
## my3High production 1.222 0.818 0.791 1.89
## factor(Parity)2 1.379 0.725 1.109 1.71
## factor(Parity)3 1.341 0.746 0.854 2.11
## DOSCC 1.124 0.890 0.917 1.38
## factor(source)US 0.782 1.279 0.460 1.33
##
## Concordance= 0.583 (se = 0.027 )
## Likelihood ratio test= 17.8 on 8 df, p=0.02
## Wald test = 82.6 on 8 df, p=1e-14
## Score (logrank) test = 18 on 8 df, p=0.02, Robust = 10.9 p=0.2
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).cm.main <- tidy(SR,conf.int = T,exp=T)[1:2,] %>% select(term,estimate,conf.low,conf.high)
cm.main <- cm.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
cm.main$production <- "All production levels"
cm.main$contrast <- "none"
cm.main$outcome <- "Clinical mastitis"Model diagnostics
SR <- cox.zph(SR)
SR## chisq df p
## Tx 3.288 2 0.19
## my3 0.369 2 0.83
## factor(Parity) 2.111 2 0.35
## DOSCC 0.940 1 0.33
## factor(source) 1.504 1 0.22
## GLOBAL 8.886 8 0.35ggcoxzph(SR)
Interaction model
CPH <- coxph(Surv(CMTAR, CM) ~ Tx*my3 + Parity + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(CMTAR, CM)
## Df Chisq Pr(>Chisq)
## Tx 2 30.51 2.4e-07 ***
## my3 2 8.02 0.018 *
## Parity 2 7.49 0.024 *
## DOSCC 1 1.27 0.259
## factor(source) 1 0.81 0.367
## Tx:my3 4 9.92 0.042 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(CPH)## Call:
## coxph(formula = Surv(CMTAR, CM) ~ Tx + my3 + Parity + DOSCC +
## factor(source) + Tx:my3, data = cow, cluster = FARMID)
##
## n= 1484, number of events= 203
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z
## TxC_SDCT -0.56604 0.56777 0.30468 0.50495 -1.12
## TxA_SDCT -0.00437 0.99564 0.29268 0.33655 -0.01
## my3Mid production 0.03085 1.03133 0.26315 0.43687 0.07
## my3High production 0.21381 1.23838 0.25637 0.38327 0.56
## Parity2 0.32306 1.38135 0.17498 0.11811 2.74
## Parity3 0.28850 1.33443 0.17974 0.23082 1.25
## DOSCC 0.12092 1.12854 0.06300 0.10724 1.13
## factor(source)US -0.24794 0.78041 0.17650 0.27466 -0.90
## TxC_SDCT:my3Mid production 0.41435 1.51339 0.41674 0.64595 0.64
## TxA_SDCT:my3Mid production -0.33488 0.71543 0.43281 0.43862 -0.76
## TxC_SDCT:my3High production 0.26960 1.30944 0.40700 0.57677 0.47
## TxA_SDCT:my3High production -0.35729 0.69957 0.43314 0.41758 -0.86
## Pr(>|z|)
## TxC_SDCT 0.2623
## TxA_SDCT 0.9896
## my3Mid production 0.9437
## my3High production 0.5769
## Parity2 0.0062 **
## Parity3 0.2113
## DOSCC 0.2595
## factor(source)US 0.3667
## TxC_SDCT:my3Mid production 0.5212
## TxA_SDCT:my3Mid production 0.4452
## TxC_SDCT:my3High production 0.6402
## TxA_SDCT:my3High production 0.3922
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 0.568 1.761 0.211 1.53
## TxA_SDCT 0.996 1.004 0.515 1.93
## my3Mid production 1.031 0.970 0.438 2.43
## my3High production 1.238 0.808 0.584 2.62
## Parity2 1.381 0.724 1.096 1.74
## Parity3 1.334 0.749 0.849 2.10
## DOSCC 1.129 0.886 0.915 1.39
## factor(source)US 0.780 1.281 0.456 1.34
## TxC_SDCT:my3Mid production 1.513 0.661 0.427 5.37
## TxA_SDCT:my3Mid production 0.715 1.398 0.303 1.69
## TxC_SDCT:my3High production 1.309 0.764 0.423 4.06
## TxA_SDCT:my3High production 0.700 1.429 0.309 1.59
##
## Concordance= 0.588 (se = 0.032 )
## Likelihood ratio test= 20.7 on 12 df, p=0.06
## Wald test = 251 on 12 df, p=<2e-16
## Score (logrank) test = 20.7 on 12 df, p=0.06, Robust = 15.3 p=0.2
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).Model diagnostics
SR <- cox.zph(CPH)
SR## chisq df p
## Tx 3.242 2 0.20
## my3 0.370 2 0.83
## Parity 2.117 2 0.35
## DOSCC 0.953 1 0.33
## factor(source) 1.507 1 0.22
## Tx:my3 4.929 4 0.29
## GLOBAL 17.342 12 0.14ggcoxzph(SR)
cow$Tx <- fct_relevel(cow$Tx,"C_SDCT","A_SDCT","BDCT")
CPH <- coxph(Surv(CMTAR, CM) ~ Tx*my3 + strata(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
out <- emmeans(CPH,pairwise ~ Tx | my3,adjust = "none",type = "response") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["ratio"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT / A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Clinical mastitis"
tab.cm <- tab
tab.cm## production contrast estimate outcome
## 1 Low production C_SDCT / BDCT 0.57 (0.21 to 1.53) Clinical mastitis
## 2 Low production A_SDCT / BDCT 0.99 (0.52 to 1.92) Clinical mastitis
## 3 Mid production C_SDCT / BDCT 0.86 (0.49 to 1.52) Clinical mastitis
## 4 Mid production A_SDCT / BDCT 0.71 (0.42 to 1.2) Clinical mastitis
## 5 High production C_SDCT / BDCT 0.74 (0.4 to 1.39) Clinical mastitis
## 6 High production A_SDCT / BDCT 0.7 (0.51 to 0.96) Clinical mastitisCulling 1-120 DIM
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
summary(KM)## Call: survfit(formula = Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
##
## Tx=C_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 1 536 1 0.998 0.00186 0.994 1.000
## 2 535 2 0.994 0.00322 0.988 1.000
## 3 533 2 0.991 0.00415 0.983 0.999
## 4 531 2 0.987 0.00490 0.977 0.997
## 5 529 1 0.985 0.00524 0.975 0.995
## 7 528 1 0.983 0.00555 0.972 0.994
## 9 527 1 0.981 0.00584 0.970 0.993
## 11 526 1 0.979 0.00612 0.968 0.992
## 13 525 1 0.978 0.00639 0.965 0.990
## 21 524 1 0.976 0.00664 0.963 0.989
## 23 523 1 0.974 0.00689 0.960 0.987
## 24 522 1 0.972 0.00712 0.958 0.986
## 33 521 1 0.970 0.00735 0.956 0.985
## 34 520 1 0.968 0.00757 0.954 0.983
## 36 519 1 0.966 0.00778 0.951 0.982
## 37 518 2 0.963 0.00819 0.947 0.979
## 38 516 1 0.961 0.00838 0.945 0.977
## 42 515 1 0.959 0.00857 0.942 0.976
## 43 514 1 0.957 0.00875 0.940 0.974
## 44 513 2 0.953 0.00911 0.936 0.971
## 45 511 1 0.951 0.00928 0.933 0.970
## 49 510 1 0.950 0.00945 0.931 0.968
## 57 509 1 0.948 0.00961 0.929 0.967
## 58 508 1 0.946 0.00977 0.927 0.965
## 63 507 1 0.944 0.00993 0.925 0.964
## 69 506 1 0.942 0.01008 0.923 0.962
## 70 505 1 0.940 0.01023 0.920 0.961
## 74 504 1 0.938 0.01038 0.918 0.959
## 75 503 1 0.937 0.01053 0.916 0.957
## 76 502 1 0.935 0.01067 0.914 0.956
## 79 501 1 0.933 0.01081 0.912 0.954
## 84 500 1 0.931 0.01095 0.910 0.953
## 89 499 1 0.929 0.01109 0.908 0.951
## 91 498 2 0.925 0.01135 0.903 0.948
## 96 496 2 0.922 0.01161 0.899 0.945
## 99 494 1 0.920 0.01173 0.897 0.943
## 111 493 2 0.916 0.01198 0.893 0.940
## 114 491 1 0.914 0.01210 0.891 0.938
## 115 490 1 0.912 0.01222 0.889 0.937
## 116 489 2 0.909 0.01245 0.885 0.933
## 117 487 2 0.905 0.01267 0.880 0.930
## 119 485 1 0.903 0.01278 0.878 0.928
##
## Tx=A_SDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 1 394 2 0.995 0.00358 0.988 1.000
## 2 392 1 0.992 0.00438 0.984 1.000
## 3 391 1 0.990 0.00505 0.980 1.000
## 4 390 1 0.987 0.00564 0.976 0.998
## 6 389 1 0.985 0.00617 0.973 0.997
## 9 388 1 0.982 0.00666 0.969 0.995
## 13 387 3 0.975 0.00792 0.959 0.990
## 18 384 1 0.972 0.00830 0.956 0.988
## 19 383 1 0.970 0.00866 0.953 0.987
## 21 382 1 0.967 0.00900 0.950 0.985
## 24 381 1 0.964 0.00933 0.946 0.983
## 33 380 1 0.962 0.00964 0.943 0.981
## 36 379 1 0.959 0.00994 0.940 0.979
## 41 378 1 0.957 0.01024 0.937 0.977
## 42 377 1 0.954 0.01052 0.934 0.975
## 44 376 1 0.952 0.01079 0.931 0.973
## 45 375 1 0.949 0.01106 0.928 0.971
## 51 374 1 0.947 0.01132 0.925 0.969
## 52 373 1 0.944 0.01157 0.922 0.967
## 54 372 1 0.942 0.01181 0.919 0.965
## 55 371 1 0.939 0.01205 0.916 0.963
## 58 370 1 0.937 0.01228 0.913 0.961
## 60 369 1 0.934 0.01251 0.910 0.959
## 62 368 1 0.931 0.01273 0.907 0.957
## 74 367 1 0.929 0.01294 0.904 0.955
## 77 366 2 0.924 0.01336 0.898 0.950
## 81 364 1 0.921 0.01356 0.895 0.948
## 85 363 1 0.919 0.01376 0.892 0.946
## 86 362 1 0.916 0.01396 0.889 0.944
## 87 361 1 0.914 0.01415 0.886 0.942
## 88 360 1 0.911 0.01433 0.884 0.940
## 90 359 1 0.909 0.01452 0.881 0.938
## 97 358 1 0.906 0.01470 0.878 0.935
## 99 357 1 0.904 0.01487 0.875 0.933
## 102 356 1 0.901 0.01505 0.872 0.931
## 105 355 1 0.898 0.01522 0.869 0.929
## 111 354 1 0.896 0.01538 0.866 0.927
## 116 353 1 0.893 0.01555 0.863 0.924
##
## Tx=BDCT
## time n.risk n.event survival std.err lower 95% CI upper 95% CI
## 3 555 1 0.998 0.00180 0.995 1.000
## 4 554 1 0.996 0.00254 0.991 1.000
## 5 553 1 0.995 0.00311 0.989 1.000
## 6 552 2 0.991 0.00401 0.983 0.999
## 7 550 2 0.987 0.00474 0.978 0.997
## 9 548 1 0.986 0.00506 0.976 0.996
## 11 547 2 0.982 0.00565 0.971 0.993
## 13 545 2 0.978 0.00617 0.966 0.991
## 15 543 1 0.977 0.00642 0.964 0.989
## 19 542 2 0.973 0.00688 0.960 0.987
## 21 540 2 0.969 0.00731 0.955 0.984
## 22 538 2 0.966 0.00772 0.951 0.981
## 23 536 2 0.962 0.00810 0.946 0.978
## 27 534 2 0.959 0.00846 0.942 0.975
## 28 532 2 0.955 0.00880 0.938 0.972
## 29 530 1 0.953 0.00897 0.936 0.971
## 30 529 1 0.951 0.00913 0.934 0.969
## 36 528 2 0.948 0.00945 0.929 0.966
## 38 526 1 0.946 0.00960 0.927 0.965
## 39 525 1 0.944 0.00975 0.925 0.963
## 41 524 2 0.941 0.01004 0.921 0.960
## 49 522 2 0.937 0.01032 0.917 0.957
## 54 520 1 0.935 0.01045 0.915 0.956
## 61 519 1 0.933 0.01059 0.913 0.954
## 64 518 1 0.932 0.01072 0.911 0.953
## 65 517 1 0.930 0.01085 0.909 0.951
## 72 516 1 0.928 0.01098 0.907 0.950
## 75 515 1 0.926 0.01110 0.905 0.948
## 78 514 1 0.924 0.01123 0.903 0.947
## 79 513 1 0.923 0.01135 0.901 0.945
## 83 512 1 0.921 0.01147 0.899 0.943
## 93 511 2 0.917 0.01170 0.894 0.940
## 99 509 2 0.914 0.01193 0.890 0.937
## 101 507 1 0.912 0.01204 0.888 0.936
## 112 506 1 0.910 0.01215 0.886 0.934
## 115 505 1 0.908 0.01226 0.884 0.932
## 116 504 1 0.906 0.01237 0.882 0.931
## 117 503 1 0.905 0.01248 0.880 0.929
## 118 502 1 0.903 0.01258 0.878 0.928ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.2),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
ylab="Cumulative incidence of \n clinical mastitis",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)
Culling - all production groups combined
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow)
km.all <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "All production levels", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Low production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="Low production"))
km.low <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "Low production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)Mid production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="Mid production"))
km.mid <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),legend.labs = c("Culture SDCT","Algorithm SDCT","Blanket DCT"),
legend = "none",
title = "Mid production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)High production cows
KM <- survfit(Surv(Cull2TAR, Cull2) ~ Tx, data = cow %>% filter(my3=="High production"))
km.high <- ggsurvplot(KM, data = cow,
pval.coord = c(0.03, 0.20),
fun = "event",xlim=c(0,120),ylim=c(0,0.25),
legend = "bottom",
title = "High production", pval = F, censor=F, conf.int = F,risk.table = F,
surv.plot.height = 5, tables.theme = theme_cleantable(),
ggtheme = theme_bw(base_family = "Times"), xlab="Days in milk",
#ylab="Cumulative incidence of \n clinical mastitis",
ylab = "",
palette = "Set1",risk.table.fontsize=3,break.time.by = 20)#png("plot.png", width=500*10, height=720*10, res=100*15)
splots <- list(km.all,
km.low,
km.mid,
km.high)
arrange_ggsurvplots(splots,ncol=2,nrow=2)
#dev.off()Cox regression
All production groups pooled together
cow$Tx <- fct_relevel(cow$Tx,"BDCT","C_SDCT","A_SDCT")
SR <- coxph(Surv(Cull2TAR, Cull2) ~ Tx + my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(SR)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 0.42 0.812
## my3 2 2.87 0.238
## factor(Parity) 2 27.43 1.1e-06 ***
## DOSCC 1 16.82 4.1e-05 ***
## factor(source) 1 3.34 0.068 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(SR)## Call:
## coxph(formula = Surv(Cull2TAR, Cull2) ~ Tx + my3 + factor(Parity) +
## DOSCC + factor(source), data = cow, cluster = FARMID)
##
## n= 1484, number of events= 148
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z Pr(>|z|)
## TxC_SDCT 0.0143 1.0144 0.1951 0.1724 0.08 0.934
## TxA_SDCT 0.0726 1.0753 0.2132 0.2183 0.33 0.739
## my3Mid production -0.2828 0.7536 0.2044 0.1849 -1.53 0.126
## my3High production -0.1157 0.8908 0.2058 0.2349 -0.49 0.622
## factor(Parity)2 0.2358 1.2660 0.2349 0.1953 1.21 0.227
## factor(Parity)3 0.9556 2.6003 0.2081 0.2116 4.52 6.3e-06 ***
## DOSCC 0.1546 1.1671 0.0717 0.0377 4.10 4.1e-05 ***
## factor(source)US 0.2910 1.3378 0.2401 0.1593 1.83 0.068 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 1.014 0.986 0.724 1.42
## TxA_SDCT 1.075 0.930 0.701 1.65
## my3Mid production 0.754 1.327 0.525 1.08
## my3High production 0.891 1.123 0.562 1.41
## factor(Parity)2 1.266 0.790 0.863 1.86
## factor(Parity)3 2.600 0.385 1.718 3.94
## DOSCC 1.167 0.857 1.084 1.26
## factor(source)US 1.338 0.747 0.979 1.83
##
## Concordance= 0.648 (se = 0.021 )
## Likelihood ratio test= 43.6 on 8 df, p=7e-07
## Wald test = 428 on 8 df, p=<2e-16
## Score (logrank) test = 46.6 on 8 df, p=2e-07, Robust = 7.97 p=0.4
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).cull.main <- tidy(SR,conf.int = T,exp=T)[1:2,] %>% select(term,estimate,conf.low,conf.high)
cull.main <- cull.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
cull.main$production <- "All production levels"
cull.main$contrast <- "none"
cull.main$outcome <- "Culling"Model diagnostics
SR <- cox.zph(SR)
SR## chisq df p
## Tx 1.442 2 0.49
## my3 3.043 2 0.22
## factor(Parity) 1.159 2 0.56
## DOSCC 0.785 1 0.38
## factor(source) 1.176 1 0.28
## GLOBAL 6.486 8 0.59ggcoxzph(SR)
Interaction model
CPH <- coxph(Surv(Cull2TAR, Cull2) ~ Tx*my3 + Parity + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 1.14 0.564
## my3 2 43.52 3.5e-10 ***
## Parity 2 27.92 8.7e-07 ***
## DOSCC 1 16.57 4.7e-05 ***
## factor(source) 1 3.01 0.083 .
## Tx:my3 4 11.33 0.023 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(CPH)## Call:
## coxph(formula = Surv(Cull2TAR, Cull2) ~ Tx + my3 + Parity + DOSCC +
## factor(source) + Tx:my3, data = cow, cluster = FARMID)
##
## n= 1484, number of events= 148
## (1 observation deleted due to missingness)
##
## coef exp(coef) se(coef) robust se z Pr(>|z|)
## TxC_SDCT 0.2839 1.3283 0.3019 0.3528 0.80 0.421
## TxA_SDCT 0.3359 1.3993 0.3214 0.2772 1.21 0.226
## my3Mid production 0.0867 1.0905 0.3284 0.3656 0.24 0.813
## my3High production 0.1032 1.1088 0.3378 0.4134 0.25 0.803
## Parity2 0.2163 1.2414 0.2352 0.1893 1.14 0.253
## Parity3 0.9361 2.5499 0.2083 0.2109 4.44 9.1e-06
## DOSCC 0.1571 1.1701 0.0717 0.0386 4.07 4.7e-05
## factor(source)US 0.2813 1.3249 0.2404 0.1620 1.74 0.083
## TxC_SDCT:my3Mid production -0.7973 0.4505 0.5005 0.5421 -1.47 0.141
## TxA_SDCT:my3Mid production -0.4060 0.6663 0.4807 0.3554 -1.14 0.253
## TxC_SDCT:my3High production -0.2283 0.7959 0.4549 0.5424 -0.42 0.674
## TxA_SDCT:my3High production -0.5517 0.5760 0.5246 0.5257 -1.05 0.294
##
## TxC_SDCT
## TxA_SDCT
## my3Mid production
## my3High production
## Parity2
## Parity3 ***
## DOSCC ***
## factor(source)US .
## TxC_SDCT:my3Mid production
## TxA_SDCT:my3Mid production
## TxC_SDCT:my3High production
## TxA_SDCT:my3High production
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## TxC_SDCT 1.328 0.753 0.665 2.65
## TxA_SDCT 1.399 0.715 0.813 2.41
## my3Mid production 1.091 0.917 0.533 2.23
## my3High production 1.109 0.902 0.493 2.49
## Parity2 1.241 0.806 0.857 1.80
## Parity3 2.550 0.392 1.686 3.86
## DOSCC 1.170 0.855 1.085 1.26
## factor(source)US 1.325 0.755 0.964 1.82
## TxC_SDCT:my3Mid production 0.451 2.220 0.156 1.30
## TxA_SDCT:my3Mid production 0.666 1.501 0.332 1.34
## TxC_SDCT:my3High production 0.796 1.256 0.275 2.30
## TxA_SDCT:my3High production 0.576 1.736 0.206 1.61
##
## Concordance= 0.655 (se = 0.018 )
## Likelihood ratio test= 47.3 on 12 df, p=4e-06
## Wald test = 3448 on 12 df, p=<2e-16
## Score (logrank) test = 50.4 on 12 df, p=1e-06, Robust = 10.4 p=0.6
##
## (Note: the likelihood ratio and score tests assume independence of
## observations within a cluster, the Wald and robust score tests do not).Model diagnostics
SR <- cox.zph(CPH)
SR## chisq df p
## Tx 1.477 2 0.48
## my3 3.045 2 0.22
## Parity 1.159 2 0.56
## DOSCC 0.766 1 0.38
## factor(source) 1.172 1 0.28
## Tx:my3 5.543 4 0.24
## GLOBAL 9.940 12 0.62ggcoxzph(SR)
cow$Tx <- fct_relevel(cow$Tx,"C_SDCT","A_SDCT","BDCT")
CPH <- coxph(Surv(Cull2TAR, Cull2) ~ Tx*my3 + factor(Parity) + DOSCC + factor(source) + cluster(FARMID), data=cow)
car::Anova(CPH)## Analysis of Deviance Table (Type II tests)
##
## Response: Surv(Cull2TAR, Cull2)
## Df Chisq Pr(>Chisq)
## Tx 2 1.14 0.564
## my3 2 43.52 3.5e-10 ***
## factor(Parity) 2 27.92 8.7e-07 ***
## DOSCC 1 16.57 4.7e-05 ***
## factor(source) 1 3.01 0.083 .
## Tx:my3 4 11.33 0.023 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1out <- emmeans(CPH,pairwise ~ Tx | my3,adjust = "none",type = "response") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["ratio"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT / A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Culling"
tab.cull <- tab
tab.cull## production contrast estimate outcome
## 1 Low production C_SDCT / BDCT 1.33 (0.67 to 2.65) Culling
## 2 Low production A_SDCT / BDCT 1.4 (0.81 to 2.41) Culling
## 3 Mid production C_SDCT / BDCT 0.6 (0.35 to 1.02) Culling
## 4 Mid production A_SDCT / BDCT 0.93 (0.57 to 1.52) Culling
## 5 High production C_SDCT / BDCT 1.06 (0.51 to 2.19) Culling
## 6 High production A_SDCT / BDCT 0.81 (0.34 to 1.92) CullingIMI post calving
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(imipc ~ Tx + my3 + Parity + factor(source) + DOSCC, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
imipc.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
imipc.main <- imipc.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
imipc.main$production <- "All production levels"
imipc.main$contrast <- "none"
imipc.main$outcome <- "IMI at calving"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(imipc ~ Tx*my3 + Parity + factor(source) + DOSCC, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = imipc ~ Tx * my3 + Parity + factor(source) +
## DOSCC, family = binomial(link = "identity"), data = qtr %>%
## filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.05659 0.05118 1.22 0.2688
## TxA_SDCT -0.03113 0.01775 3.07 0.0796 .
## TxBDCT -0.02591 0.03550 0.53 0.4655
## my3Mid production 0.00515 0.02419 0.05 0.8315
## my3High production 0.00419 0.02207 0.04 0.8493
## Parity2 0.00789 0.02127 0.14 0.7108
## Parity3 0.01872 0.01725 1.18 0.2778
## factor(source)US 0.06181 0.04640 1.78 0.1828
## DOSCC 0.02627 0.00892 8.68 0.0032 **
## TxA_SDCT:my3Mid production 0.00629 0.04122 0.02 0.8787
## TxBDCT:my3Mid production -0.00619 0.03319 0.03 0.8521
## TxA_SDCT:my3High production 0.02749 0.03016 0.83 0.3620
## TxBDCT:my3High production 0.01395 0.03543 0.16 0.6938
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.143
## Number of clusters: 16 Maximum cluster size: 1270emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.213 0.0270 Inf 0.1597 0.266
## A_SDCT Low production 0.181 0.0273 Inf 0.1280 0.235
## BDCT Low production 0.187 0.0419 Inf 0.1045 0.269
## C_SDCT Mid production 0.218 0.0270 Inf 0.1648 0.271
## A_SDCT Mid production 0.193 0.0249 Inf 0.1440 0.242
## BDCT Mid production 0.186 0.0445 Inf 0.0983 0.273
## C_SDCT High production 0.217 0.0191 Inf 0.1794 0.254
## A_SDCT High production 0.213 0.0219 Inf 0.1701 0.256
## BDCT High production 0.205 0.0200 Inf 0.1656 0.244
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
tab$outcome <- "IMI at calving"
tab.imipc <- tab
tab.imipc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.03 (-0.04 to 0.1) IMI at calving
## 2 Low production A_SDCT - BDCT -0.01 (-0.05 to 0.04) IMI at calving
## 3 Mid production C_SDCT - BDCT 0.03 (-0.03 to 0.1) IMI at calving
## 4 Mid production A_SDCT - BDCT 0.01 (-0.08 to 0.1) IMI at calving
## 5 High production C_SDCT - BDCT 0.01 (-0.03 to 0.05) IMI at calving
## 6 High production A_SDCT - BDCT 0.01 (-0.03 to 0.04) IMI at calvingtab.imipc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.03 (-0.04 to 0.1) IMI at calving
## 2 Low production A_SDCT - BDCT -0.01 (-0.05 to 0.04) IMI at calving
## 3 Mid production C_SDCT - BDCT 0.03 (-0.03 to 0.1) IMI at calving
## 4 Mid production A_SDCT - BDCT 0.01 (-0.08 to 0.1) IMI at calving
## 5 High production C_SDCT - BDCT 0.01 (-0.03 to 0.05) IMI at calving
## 6 High production A_SDCT - BDCT 0.01 (-0.03 to 0.04) IMI at calvingUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(imipc ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(imipc ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: imipc ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 5110 5201 -2541 5082 5079
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -0.891 -0.602 -0.432 -0.242 4.778
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.392 0.626
## Number of obs: 5093, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.0413 0.2964 -6.89 5.7e-12 ***
## TxA_SDCT -0.1741 0.1574 -1.11 0.2685
## TxBDCT -0.0821 0.1394 -0.59 0.5559
## my3Mid production 0.0195 0.1445 0.13 0.8929
## my3High production 0.0172 0.1434 0.12 0.9044
## Parity2 0.1661 0.0869 1.91 0.0561 .
## Parity3 0.2447 0.0894 2.74 0.0062 **
## DOSCC 0.0837 0.0335 2.50 0.0125 *
## factor(source)US -0.1503 0.3374 -0.45 0.6559
## TxA_SDCT:my3Mid production 0.1049 0.2163 0.49 0.6276
## TxBDCT:my3Mid production -0.0806 0.2018 -0.40 0.6896
## TxA_SDCT:my3High production 0.1210 0.2191 0.55 0.5807
## TxBDCT:my3High production 0.0283 0.1992 0.14 0.8871
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: imipc
## Chisq Df Pr(>Chisq)
## Tx 1.76 2 0.414
## my3 0.44 2 0.805
## Parity 8.03 2 0.018 *
## DOSCC 6.24 1 0.013 *
## factor(source) 0.20 1 0.656
## Tx:my3 0.91 4 0.922
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: imipc ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 5292
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.0186 -0.6560 -0.4066 -0.0993 2.4722
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.0078 0.0883
## Residual 0.1621 0.4026
## Number of obs: 5093, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.10875 0.04478 2.43
## TxA_SDCT -0.02940 0.02600 -1.13
## TxBDCT -0.01322 0.02289 -0.58
## my3Mid production 0.00334 0.02400 0.14
## my3High production 0.00324 0.02367 0.14
## Parity2 0.02561 0.01415 1.81
## Parity3 0.03888 0.01471 2.64
## DOSCC 0.01379 0.00557 2.47
## factor(source)US -0.00573 0.04765 -0.12
## TxA_SDCT:my3Mid production 0.01755 0.03546 0.49
## TxBDCT:my3Mid production -0.01334 0.03279 -0.41
## TxA_SDCT:my3High production 0.02002 0.03599 0.56
## TxBDCT:my3High production 0.00390 0.03237 0.12car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: imipc
## Chisq Df Pr(>Chisq)
## Tx 1.84 2 0.399
## my3 0.46 2 0.795
## Parity 7.49 2 0.024 *
## DOSCC 6.12 1 0.013 *
## factor(source) 0.01 1 0.904
## Tx:my3 0.96 4 0.916
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.167 0.0270 Inf 0.120 0.226
## A_SDCT Low production 0.144 0.0255 Inf 0.101 0.201
## BDCT Low production 0.156 0.0249 Inf 0.113 0.211
## C_SDCT Mid production 0.169 0.0271 Inf 0.123 0.229
## A_SDCT Mid production 0.160 0.0272 Inf 0.114 0.221
## BDCT Mid production 0.148 0.0246 Inf 0.106 0.203
## C_SDCT High production 0.169 0.0268 Inf 0.123 0.228
## A_SDCT High production 0.162 0.0282 Inf 0.114 0.225
## BDCT High production 0.162 0.0263 Inf 0.117 0.220
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.01 (-0.03 to 0.06)
## 2 Low production A_SDCT - BDCT -0.02 (-0.06 to 0.03)
## 3 Mid production C_SDCT - BDCT 0.03 (-0.02 to 0.07)
## 4 Mid production A_SDCT - BDCT 0.01 (-0.03 to 0.06)
## 5 High production C_SDCT - BDCT 0.01 (-0.04 to 0.05)
## 6 High production A_SDCT - BDCT 0 (-0.05 to 0.05)IMI cure
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
qtr <- qtr %>% arrange(FARMID,CowID)
gee <- geeglm(cure ~ Tx + my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
imicure.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
imicure.main <- imicure.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
imicure.main$production <- "All production levels"
imicure.main$contrast <- "none"
imicure.main$outcome <- "IMI cure"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
gee <- geeglm(cure ~ Tx*my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ Tx * my3 + Parity + DOSCC + factor(source),
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(cure) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 1.027043 0.033706 928.47 < 2e-16 ***
## TxA_SDCT -0.015212 0.019302 0.62 0.43064
## TxBDCT -0.037237 0.031191 1.43 0.23255
## my3Mid production -0.032254 0.021755 2.20 0.13817
## my3High production 0.000819 0.019762 0.00 0.96696
## Parity2 -0.032772 0.014248 5.29 0.02144 *
## Parity3 -0.034443 0.029314 1.38 0.24000
## DOSCC -0.015560 0.005848 7.08 0.00780 **
## factor(source)US -0.048110 0.031144 2.39 0.12241
## TxA_SDCT:my3Mid production 0.080852 0.021165 14.59 0.00013 ***
## TxBDCT:my3Mid production 0.053197 0.035965 2.19 0.13910
## TxA_SDCT:my3High production -0.022433 0.051764 0.19 0.66474
## TxBDCT:my3High production -0.005079 0.029420 0.03 0.86295
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.327
## Number of clusters: 16 Maximum cluster size: 432emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.909 0.0248 Inf 0.860 0.958
## A_SDCT Low production 0.894 0.0278 Inf 0.839 0.948
## BDCT Low production 0.872 0.0332 Inf 0.807 0.937
## C_SDCT Mid production 0.877 0.0231 Inf 0.831 0.922
## A_SDCT Mid production 0.942 0.0173 Inf 0.908 0.976
## BDCT Mid production 0.893 0.0289 Inf 0.836 0.949
## C_SDCT High production 0.910 0.0122 Inf 0.886 0.934
## A_SDCT High production 0.872 0.0382 Inf 0.797 0.947
## BDCT High production 0.867 0.0152 Inf 0.838 0.897
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
tab$outcome <- "IMI cure"
tab.imicure <- tab
tab.imicure## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.04 (-0.02 to 0.1) IMI cure
## 2 Low production A_SDCT - BDCT 0.02 (-0.03 to 0.07) IMI cure
## 3 Mid production C_SDCT - BDCT -0.02 (-0.07 to 0.04) IMI cure
## 4 Mid production A_SDCT - BDCT 0.05 (0.01 to 0.09) IMI cure
## 5 High production C_SDCT - BDCT 0.04 (0.01 to 0.08) IMI cure
## 6 High production A_SDCT - BDCT 0 (-0.08 to 0.09) IMI cureUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(cure ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(cure ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: cure ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 874 946 -423 846 1171
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.524 0.254 0.331 0.410 0.590
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.317 0.563
## Number of obs: 1185, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 3.0833 0.6303 4.89 1e-06 ***
## TxA_SDCT -0.1196 0.4231 -0.28 0.777
## TxBDCT -0.1995 0.3621 -0.55 0.582
## my3Mid production -0.2094 0.3807 -0.55 0.582
## my3High production 0.1331 0.3995 0.33 0.739
## Parity2 -0.4134 0.2230 -1.85 0.064 .
## Parity3 -0.2413 0.2362 -1.02 0.307
## DOSCC -0.0819 0.0900 -0.91 0.363
## factor(source)US -0.0877 0.4114 -0.21 0.831
## TxA_SDCT:my3Mid production 0.7015 0.5958 1.18 0.239
## TxBDCT:my3Mid production 0.5562 0.5347 1.04 0.298
## TxA_SDCT:my3High production -0.2692 0.5852 -0.46 0.646
## TxBDCT:my3High production -0.1806 0.5157 -0.35 0.726
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## optimizer (Nelder_Mead) convergence code: 0 (OK)
## Model failed to converge with max|grad| = 0.00795887 (tol = 0.002, component 1)car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: cure
## Chisq Df Pr(>Chisq)
## Tx 0.21 2 0.90
## my3 0.63 2 0.73
## Parity 3.46 2 0.18
## DOSCC 0.83 1 0.36
## factor(source) 0.05 1 0.83
## Tx:my3 3.55 4 0.47summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: cure ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 747
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.995 0.190 0.329 0.466 0.691
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.00278 0.0527
## Residual 0.10364 0.3219
## Number of obs: 1185, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.96860 0.06142 15.77
## TxA_SDCT -0.01222 0.04514 -0.27
## TxBDCT -0.02100 0.03810 -0.55
## my3Mid production -0.02020 0.03958 -0.51
## my3High production 0.01263 0.03955 0.32
## Parity2 -0.04005 0.02298 -1.74
## Parity3 -0.02390 0.02413 -0.99
## DOSCC -0.00785 0.00931 -0.84
## factor(source)US -0.00449 0.03790 -0.12
## TxA_SDCT:my3Mid production 0.06622 0.06034 1.10
## TxBDCT:my3Mid production 0.05478 0.05449 1.01
## TxA_SDCT:my3High production -0.02652 0.06132 -0.43
## TxBDCT:my3High production -0.01725 0.05340 -0.32car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: cure
## Chisq Df Pr(>Chisq)
## Tx 0.27 2 0.87
## my3 0.74 2 0.69
## Parity 3.11 2 0.21
## DOSCC 0.71 1 0.40
## factor(source) 0.01 1 0.91
## Tx:my3 3.29 4 0.51emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.920 0.0256 Inf 0.853 0.958
## A_SDCT Low production 0.911 0.0314 Inf 0.827 0.956
## BDCT Low production 0.904 0.0268 Inf 0.837 0.945
## C_SDCT Mid production 0.903 0.0281 Inf 0.833 0.946
## A_SDCT Mid production 0.944 0.0217 Inf 0.883 0.974
## BDCT Mid production 0.930 0.0237 Inf 0.867 0.965
## C_SDCT High production 0.929 0.0227 Inf 0.870 0.963
## A_SDCT High production 0.899 0.0339 Inf 0.811 0.949
## BDCT High production 0.900 0.0283 Inf 0.829 0.943
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["lower.CL"]],
ucl = out[["contrasts"]][["upper.CL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.02 (-0.05 to 0.1)
## 2 Low production A_SDCT - BDCT 0.01 (-0.08 to 0.09)
## 3 Mid production C_SDCT - BDCT -0.03 (-0.11 to 0.04)
## 4 Mid production A_SDCT - BDCT 0.02 (-0.06 to 0.1)
## 5 High production C_SDCT - BDCT 0.04 (-0.03 to 0.11)
## 6 High production A_SDCT - BDCT 0 (-0.08 to 0.08)NewIMI
All production groups pooled together
qtr$Tx <- fct_relevel(qtr$Tx,"BDCT","C_SDCT","A_SDCT")
gee <- geeglm(newimi ~ Tx + my3 + Parity + DOSCC + factor(source), data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
iminew.main <- tidy(gee,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
iminew.main <- iminew.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
iminew.main$production <- "All production levels"
iminew.main$contrast <- "none"
iminew.main$outcome <- "New IMI risk"Interaction model
qtr$Tx <- fct_relevel(qtr$Tx,"C_SDCT","A_SDCT","BDCT")
gee <- geeglm(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) , data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ Tx * my3 + Parity + DOSCC + factor(source),
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(newimi) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.04878 0.04614 1.12 0.29040
## TxA_SDCT -0.02134 0.02341 0.83 0.36193
## TxBDCT -0.02619 0.03748 0.49 0.48468
## my3Mid production 0.00165 0.02321 0.01 0.94332
## my3High production 0.00701 0.02263 0.10 0.75662
## Parity2 0.00749 0.02219 0.11 0.73578
## Parity3 0.01620 0.01780 0.83 0.36290
## DOSCC 0.02606 0.00778 11.23 0.00081 ***
## factor(source)US 0.06057 0.05005 1.46 0.22620
## TxA_SDCT:my3Mid production -0.00260 0.03777 0.00 0.94513
## TxBDCT:my3Mid production 0.00034 0.03382 0.00 0.99198
## TxA_SDCT:my3High production 0.00558 0.02465 0.05 0.82085
## TxBDCT:my3High production 0.00280 0.03877 0.01 0.94239
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.166
## Number of clusters: 16 Maximum cluster size: 1223emmeans(gee, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 emmean SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.203 0.0290 Inf 0.1459 0.260
## A_SDCT Low production 0.181 0.0330 Inf 0.1168 0.246
## BDCT Low production 0.177 0.0451 Inf 0.0881 0.265
## C_SDCT Mid production 0.204 0.0270 Inf 0.1515 0.257
## A_SDCT Mid production 0.180 0.0244 Inf 0.1327 0.228
## BDCT Mid production 0.179 0.0481 Inf 0.0843 0.273
## C_SDCT High production 0.210 0.0202 Inf 0.1702 0.249
## A_SDCT High production 0.194 0.0208 Inf 0.1532 0.235
## BDCT High production 0.186 0.0214 Inf 0.1445 0.228
##
## Results are averaged over the levels of: Parity, source
## Covariance estimate used: vbeta
## Confidence level used: 0.95out <- emmeans(gee,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
mean <- data.frame(
production = out[["emmeans"]][["my3"]],
tx = out[["emmeans"]][["Tx"]],
est = out[["emmeans"]][["emmean"]]
)
mean$est <- mean$est * 100
#mean
tab$outcome <- "New IMI risk"
tab.newimi <- tab
tab.newimi## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.03 (-0.05 to 0.1) New IMI risk
## 2 Low production A_SDCT - BDCT 0 (-0.04 to 0.05) New IMI risk
## 3 Mid production C_SDCT - BDCT 0.03 (-0.04 to 0.1) New IMI risk
## 4 Mid production A_SDCT - BDCT 0 (-0.09 to 0.09) New IMI risk
## 5 High production C_SDCT - BDCT 0.02 (-0.01 to 0.06) New IMI risk
## 6 High production A_SDCT - BDCT 0.01 (-0.03 to 0.04) New IMI riskUsing GLMM (logistic) and LMM (i.e. to estimate risk difference)
glmm <- glmer(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr, family=binomial())
lmm <- lmer(newimi ~ Tx*my3 + Parity + DOSCC + factor(source) + (1|FARMID), data=qtr)
summary(glmm)## Generalized linear mixed model fit by maximum likelihood (Laplace
## Approximation) [glmerMod]
## Family: binomial ( logit )
## Formula: newimi ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## AIC BIC logLik deviance df.resid
## 4486 4576 -2229 4458 4620
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -0.867 -0.592 -0.400 -0.236 5.005
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.419 0.647
## Number of obs: 4634, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.07924 0.30937 -6.72 1.8e-11 ***
## TxA_SDCT -0.15138 0.16842 -0.90 0.3687
## TxBDCT -0.08989 0.14871 -0.60 0.5455
## my3Mid production 0.00107 0.15416 0.01 0.9945
## my3High production 0.02293 0.15263 0.15 0.8806
## Parity2 0.19924 0.09300 2.14 0.0322 *
## Parity3 0.26461 0.09565 2.77 0.0057 **
## DOSCC 0.07700 0.03521 2.19 0.0288 *
## factor(source)US -0.18850 0.34989 -0.54 0.5901
## TxA_SDCT:my3Mid production 0.08260 0.23066 0.36 0.7203
## TxBDCT:my3Mid production -0.04536 0.21507 -0.21 0.8330
## TxA_SDCT:my3High production 0.02681 0.23619 0.11 0.9096
## TxBDCT:my3High production -0.04300 0.21474 -0.20 0.8413
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## optimizer (Nelder_Mead) convergence code: 0 (OK)
## Model failed to converge with max|grad| = 0.00234589 (tol = 0.002, component 1)car::Anova(glmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: newimi
## Chisq Df Pr(>Chisq)
## Tx 2.23 2 0.329
## my3 0.02 2 0.989
## Parity 8.66 2 0.013 *
## DOSCC 4.78 1 0.029 *
## factor(source) 0.29 1 0.590
## Tx:my3 0.33 4 0.988
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: newimi ~ Tx * my3 + Parity + DOSCC + factor(source) + (1 | FARMID)
## Data: qtr
##
## REML criterion at convergence: 4612
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.005 -0.658 -0.372 -0.114 2.528
##
## Random effects:
## Groups Name Variance Std.Dev.
## FARMID (Intercept) 0.00784 0.0886
## Residual 0.15489 0.3936
## Number of obs: 4634, groups: FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 0.108372 0.045094 2.40
## TxA_SDCT -0.024427 0.026839 -0.91
## TxBDCT -0.013818 0.023383 -0.59
## my3Mid production 0.000216 0.024524 0.01
## my3High production 0.003869 0.024161 0.16
## Parity2 0.029519 0.014482 2.04
## Parity3 0.040121 0.015054 2.67
## DOSCC 0.012050 0.005622 2.14
## factor(source)US -0.009240 0.047793 -0.19
## TxA_SDCT:my3Mid production 0.013322 0.036304 0.37
## TxBDCT:my3Mid production -0.007517 0.033498 -0.22
## TxA_SDCT:my3High production 0.003706 0.037180 0.10
## TxBDCT:my3High production -0.006786 0.033209 -0.20car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: newimi
## Chisq Df Pr(>Chisq)
## Tx 2.29 2 0.317
## my3 0.02 2 0.989
## Parity 8.05 2 0.018 *
## DOSCC 4.59 1 0.032 *
## factor(source) 0.04 1 0.847
## Tx:my3 0.36 4 0.985
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1emmeans(glmm, ~ Tx*my3,type = "response",adjust = "none")## Tx my3 prob SE df asymp.LCL asymp.UCL
## C_SDCT Low production 0.158 0.0268 Inf 0.1117 0.217
## A_SDCT Low production 0.139 0.0259 Inf 0.0951 0.197
## BDCT Low production 0.146 0.0247 Inf 0.1039 0.201
## C_SDCT Mid production 0.158 0.0267 Inf 0.1120 0.217
## A_SDCT Mid production 0.149 0.0267 Inf 0.1036 0.209
## BDCT Mid production 0.141 0.0245 Inf 0.0989 0.196
## C_SDCT High production 0.161 0.0269 Inf 0.1146 0.220
## A_SDCT High production 0.144 0.0272 Inf 0.0989 0.206
## BDCT High production 0.143 0.0251 Inf 0.1009 0.200
##
## Results are averaged over the levels of: Parity, source
## Confidence level used: 0.95
## Intervals are back-transformed from the logit scaleout <- emmeans(glmm,pairwise ~ Tx | my3,adjust = "none",type="response") %>% confint
out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.01 (-0.03 to 0.06)
## 2 Low production A_SDCT - BDCT -0.01 (-0.06 to 0.04)
## 3 Mid production C_SDCT - BDCT 0.02 (-0.03 to 0.07)
## 4 Mid production A_SDCT - BDCT 0.01 (-0.04 to 0.06)
## 5 High production C_SDCT - BDCT 0.02 (-0.03 to 0.07)
## 6 High production A_SDCT - BDCT 0 (-0.05 to 0.05)DHI Dataset
Somatic cell count (1-120 DIM)
All production groups pooled together
dhi$Tx <- fct_relevel(dhi$Tx,"BDCT","C_SDCT","A_SDCT")
dhi <- dhi %>% arrange(FARMID,CowID)
mm0 <- geeglm(scc ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID),
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID), data = dhi %>% filter(!is.na(DOSCC)), id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.32311 0.18494 322.86 < 2e-16 ***
## TxC_SDCT 0.02880 0.06400 0.20 0.6527
## TxA_SDCT 0.02785 0.06893 0.16 0.6862
## my3Mid production 0.00323 0.06804 0.00 0.9622
## my3High production 0.03985 0.06938 0.33 0.5657
## TestDIMcat2030 -0.54180 0.06464 70.25 < 2e-16 ***
## TestDIMcat2050 -0.55646 0.05901 88.92 < 2e-16 ***
## TestDIMcat2070 -0.41845 0.06065 47.61 5.2e-12 ***
## TestDIMcat2090 -0.28694 0.06479 19.61 9.5e-06 ***
## TestDIMcat20110 -0.12394 0.06105 4.12 0.0423 *
## DOSCC 0.17325 0.02696 41.29 1.3e-10 ***
## Parity2 0.13258 0.06260 4.49 0.0342 *
## Parity3 0.30245 0.07346 16.95 3.8e-05 ***
## factor(FARMID)2 0.07295 0.16932 0.19 0.6666
## factor(FARMID)3 -0.01979 0.13236 0.02 0.8811
## factor(FARMID)4 0.06333 0.12810 0.24 0.6210
## factor(FARMID)5 0.03105 0.13571 0.05 0.8190
## factor(FARMID)6 -0.47480 0.17696 7.20 0.0073 **
## factor(FARMID)7 -0.27231 0.14619 3.47 0.0625 .
## factor(FARMID)8 -0.07255 0.27972 0.07 0.7954
## factor(FARMID)9 -0.06666 0.29768 0.05 0.8228
## factor(FARMID)10 -0.24485 0.22459 1.19 0.2756
## factor(FARMID)11 -0.40008 0.23970 2.79 0.0951 .
## factor(FARMID)12 -0.47358 0.21677 4.77 0.0289 *
## factor(FARMID)13 0.11772 0.30299 0.15 0.6976
## factor(FARMID)14 0.10451 0.25583 0.17 0.6829
## factor(FARMID)15 0.03146 0.37048 0.01 0.9323
## factor(FARMID)16 -0.05681 0.23297 0.06 0.8074
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0567
## Number of clusters: 1364 Maximum cluster size: 4scc.main <- tidy(mm0,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
scc.main <- scc.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
scc.main$production <- "All production levels"
scc.main$contrast <- "none"
scc.main$outcome <- "Log SCC (1-120 DIM)"atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(Tx,TestDIMcat20,SCC,LCL,UCL))
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=Tx, colour=Tx) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$LCL, ymax=emm$UCL,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve
Interaction model (Tx * milk yield)
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(scc ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID),
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID), data = dhi %>% filter(!is.na(DOSCC)), id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.2182 0.1961 269.20 < 2e-16 ***
## TxA_SDCT 0.2199 0.1235 3.17 0.0751 .
## TxBDCT 0.1150 0.1033 1.24 0.2657
## my3Mid production 0.1954 0.1145 2.91 0.0879 .
## my3High production 0.1936 0.1081 3.21 0.0733 .
## TestDIMcat2030 -0.5369 0.0647 68.93 < 2e-16 ***
## TestDIMcat2050 -0.5550 0.0591 88.29 < 2e-16 ***
## TestDIMcat2070 -0.4130 0.0605 46.59 8.7e-12 ***
## TestDIMcat2090 -0.2844 0.0650 19.17 1.2e-05 ***
## TestDIMcat20110 -0.1197 0.0611 3.85 0.0499 *
## DOSCC 0.1761 0.0271 42.18 8.3e-11 ***
## Parity2 0.1377 0.0627 4.82 0.0281 *
## Parity3 0.3037 0.0734 17.11 3.5e-05 ***
## factor(FARMID)2 0.0711 0.1683 0.18 0.6726
## factor(FARMID)3 -0.0230 0.1321 0.03 0.8617
## factor(FARMID)4 0.0610 0.1280 0.23 0.6334
## factor(FARMID)5 0.0232 0.1354 0.03 0.8641
## factor(FARMID)6 -0.4867 0.1775 7.52 0.0061 **
## factor(FARMID)7 -0.2762 0.1464 3.56 0.0592 .
## factor(FARMID)8 -0.0840 0.2774 0.09 0.7620
## factor(FARMID)9 -0.0626 0.2978 0.04 0.8334
## factor(FARMID)10 -0.2720 0.2241 1.47 0.2247
## factor(FARMID)11 -0.3806 0.2441 2.43 0.1189
## factor(FARMID)12 -0.4640 0.2189 4.50 0.0340 *
## factor(FARMID)13 0.1154 0.3017 0.15 0.7022
## factor(FARMID)14 0.1116 0.2549 0.19 0.6616
## factor(FARMID)15 0.0447 0.3672 0.01 0.9031
## factor(FARMID)16 -0.0612 0.2334 0.07 0.7932
## TxA_SDCT:my3Mid production -0.3902 0.1646 5.62 0.0178 *
## TxBDCT:my3Mid production -0.1999 0.1581 1.60 0.2060
## TxA_SDCT:my3High production -0.2424 0.1699 2.03 0.1538
## TxBDCT:my3High production -0.2261 0.1502 2.26 0.1324
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0562
## Number of clusters: 1364 Maximum cluster size: 4out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Log SCC (1-120 DIM)"
tab.scc <- tab
tab.scc## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT -0.12 (-0.32 to 0.09) Log SCC (1-120 DIM)
## 2 Low production A_SDCT - BDCT 0.1 (-0.13 to 0.34) Log SCC (1-120 DIM)
## 3 Mid production C_SDCT - BDCT 0.08 (-0.15 to 0.32) Log SCC (1-120 DIM)
## 4 Mid production A_SDCT - BDCT -0.09 (-0.3 to 0.13) Log SCC (1-120 DIM)
## 5 High production C_SDCT - BDCT 0.11 (-0.1 to 0.32) Log SCC (1-120 DIM)
## 6 High production A_SDCT - BDCT 0.09 (-0.15 to 0.33) Log SCC (1-120 DIM)atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(my3,Tx,TestDIMcat20,SCC,LCL,UCL))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=Tx, colour=Tx) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$LCL, ymax=emm$UCL,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1") + facet_grid( ~ my3)
curve
Interaction model using linear mixed model
lmm <- lmer(scc ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + (1|FARMID/CowID),
data=dhi)
summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: scc ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity + (1 | FARMID/CowID)
## Data: dhi
##
## REML criterion at convergence: 14781
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.123 -0.522 -0.107 0.439 4.124
##
## Random effects:
## Groups Name Variance Std.Dev.
## CowID:FARMID (Intercept) 0.6998 0.837
## FARMID (Intercept) 0.0127 0.113
## Residual 1.0526 1.026
## Number of obs: 4553, groups: CowID:FARMID, 1364; FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 3.1076 0.1554 20.00
## TxA_SDCT 0.2463 0.1226 2.01
## TxBDCT 0.1192 0.1108 1.08
## my3Mid production 0.1914 0.1154 1.66
## my3High production 0.2211 0.1138 1.94
## TestDIMcat2030 -0.5653 0.0576 -9.81
## TestDIMcat2050 -0.5733 0.0552 -10.39
## TestDIMcat2070 -0.4171 0.0548 -7.61
## TestDIMcat2090 -0.2994 0.0586 -5.11
## TestDIMcat20110 -0.1163 0.0548 -2.12
## DOSCC 0.1876 0.0261 7.20
## Parity2 0.1102 0.0684 1.61
## Parity3 0.2989 0.0712 4.20
## TxA_SDCT:my3Mid production -0.3728 0.1681 -2.22
## TxBDCT:my3Mid production -0.1798 0.1578 -1.14
## TxA_SDCT:my3High production -0.2410 0.1713 -1.41
## TxBDCT:my3High production -0.2635 0.1576 -1.67car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: scc
## Chisq Df Pr(>Chisq)
## Tx 0.81 2 0.66624
## my3 0.69 2 0.70905
## TestDIMcat20 183.11 5 < 2e-16 ***
## DOSCC 51.83 1 6.1e-13 ***
## Parity 17.68 2 0.00014 ***
## Tx:my3 6.54 4 0.16232
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT -0.12 (-0.34 to 0.1)
## 2 Low production A_SDCT - BDCT 0.13 (-0.1 to 0.36)
## 3 Mid production C_SDCT - BDCT 0.06 (-0.16 to 0.28)
## 4 Mid production A_SDCT - BDCT -0.07 (-0.3 to 0.16)
## 5 High production C_SDCT - BDCT 0.14 (-0.07 to 0.36)
## 6 High production A_SDCT - BDCT 0.15 (-0.1 to 0.39)Milk Yield (1-120 DIM)
All production groups pooled together
dhi$Tx <- fct_relevel(dhi$Tx,"BDCT","C_SDCT","A_SDCT")
dhi <- dhi %>% arrange(FARMID,CowID)
mm0 <- geeglm(MY ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID) + DOMY,
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx + my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID) + DOMY, data = dhi %>% filter(!is.na(DOSCC)),
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 32.38133 1.51991 453.90 < 2e-16 ***
## TxC_SDCT -0.20428 0.41036 0.25 0.6186
## TxA_SDCT -0.95448 0.45459 4.41 0.0358 *
## my3Mid production -0.00291 0.55694 0.00 0.9958
## my3High production 1.68138 0.81346 4.27 0.0387 *
## TestDIMcat2030 10.64886 0.45885 538.60 < 2e-16 ***
## TestDIMcat2050 12.29294 0.40042 942.50 < 2e-16 ***
## TestDIMcat2070 11.57778 0.42798 731.80 < 2e-16 ***
## TestDIMcat2090 10.09220 0.44159 522.33 < 2e-16 ***
## TestDIMcat20110 7.94226 0.43717 330.05 < 2e-16 ***
## DOSCC 0.15269 0.17494 0.76 0.3827
## Parity2 2.91192 0.42390 47.19 6.4e-12 ***
## Parity3 3.39658 0.49713 46.68 8.4e-12 ***
## factor(FARMID)2 -1.19254 1.05352 1.28 0.2577
## factor(FARMID)3 -4.59538 0.88976 26.67 2.4e-07 ***
## factor(FARMID)4 5.47511 0.91717 35.64 2.4e-09 ***
## factor(FARMID)5 0.65044 0.90301 0.52 0.4713
## factor(FARMID)6 2.36346 1.07481 4.84 0.0279 *
## factor(FARMID)7 2.23025 1.03413 4.65 0.0310 *
## factor(FARMID)8 -2.05747 1.63115 1.59 0.2072
## factor(FARMID)9 -7.93355 1.67387 22.46 2.1e-06 ***
## factor(FARMID)10 -1.56270 1.23962 1.59 0.2074
## factor(FARMID)11 -3.03963 1.70123 3.19 0.0740 .
## factor(FARMID)12 -4.23579 1.71188 6.12 0.0133 *
## factor(FARMID)13 -4.97843 1.28275 15.06 0.0001 ***
## factor(FARMID)14 -2.02688 2.43435 0.69 0.4051
## factor(FARMID)15 -4.83048 1.70948 7.98 0.0047 **
## factor(FARMID)16 -2.35627 1.22689 3.69 0.0548 .
## DOMY 0.13808 0.04720 8.56 0.0034 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 71.1 2.37
## Number of clusters: 1363 Maximum cluster size: 4my.main <- tidy(mm0,conf.int = T)[2:3,] %>% select(term,estimate,conf.low,conf.high)
my.main <- my.main %>% mutate(
estimate = paste0(round(estimate,2),
" (",
round(conf.low,2),
" to ",
round(conf.high,2),
")"),
conf.low = NULL,
conf.high = NULL)
my.main$production <- "All production levels"
my.main$contrast <- "none"
my.main$outcome <- "Milk Yield (1-120 DIM)"atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm <- emm %>% subset(select=c(Tx,TestDIMcat20,estimate,conf.low,conf.high))
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(30,55))) + aes(x=as.numeric(TestDIMcat20), y=estimate, group=Tx, colour=Tx) +
labs(y = "Milk Yield)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$conf.low, ymax=emm$conf.high,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve
Interaction model (Tx * milk yield)
dhi$Tx <- fct_relevel(dhi$Tx,"C_SDCT","A_SDCT","BDCT")
mm0 <- geeglm(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + factor(FARMID) + DOMY,
data=dhi %>% filter(!is.na(DOSCC))
,id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity +
## factor(FARMID) + DOMY, data = dhi %>% filter(!is.na(DOSCC)),
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 32.6753 1.5311 455.45 < 2e-16 ***
## TxA_SDCT -1.5043 0.8398 3.21 0.0732 .
## TxBDCT -0.3414 0.7080 0.23 0.6296
## my3Mid production -0.3237 0.8131 0.16 0.6905
## my3High production 0.8216 1.0734 0.59 0.4441
## TestDIMcat2030 10.6440 0.4580 540.07 < 2e-16 ***
## TestDIMcat2050 12.2954 0.4001 944.43 < 2e-16 ***
## TestDIMcat2070 11.5710 0.4271 734.03 < 2e-16 ***
## TestDIMcat2090 10.0872 0.4414 522.35 < 2e-16 ***
## TestDIMcat20110 7.9326 0.4366 330.05 < 2e-16 ***
## DOSCC 0.1420 0.1743 0.66 0.4151
## Parity2 2.8950 0.4226 46.92 7.4e-12 ***
## Parity3 3.4053 0.4951 47.30 6.1e-12 ***
## factor(FARMID)2 -1.1698 1.0525 1.24 0.2664
## factor(FARMID)3 -4.6148 0.8890 26.95 2.1e-07 ***
## factor(FARMID)4 5.4795 0.9171 35.70 2.3e-09 ***
## factor(FARMID)5 0.6655 0.9011 0.55 0.4602
## factor(FARMID)6 2.4130 1.0759 5.03 0.0249 *
## factor(FARMID)7 2.2174 1.0331 4.61 0.0318 *
## factor(FARMID)8 -2.0777 1.6152 1.65 0.1983
## factor(FARMID)9 -7.9371 1.6747 22.46 2.1e-06 ***
## factor(FARMID)10 -1.4768 1.2438 1.41 0.2351
## factor(FARMID)11 -3.2404 1.6901 3.68 0.0552 .
## factor(FARMID)12 -4.3655 1.6892 6.68 0.0098 **
## factor(FARMID)13 -4.9717 1.2759 15.18 9.8e-05 ***
## factor(FARMID)14 -1.9944 2.4385 0.67 0.4134
## factor(FARMID)15 -4.8663 1.6905 8.29 0.0040 **
## factor(FARMID)16 -2.2892 1.2321 3.45 0.0632 .
## DOMY 0.1373 0.0474 8.37 0.0038 **
## TxA_SDCT:my3Mid production 0.9239 1.0880 0.72 0.3958
## TxBDCT:my3Mid production 0.0794 0.9563 0.01 0.9338
## TxA_SDCT:my3High production 1.1998 1.1562 1.08 0.2994
## TxBDCT:my3High production 1.5158 1.0321 2.16 0.1419
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 71 2.37
## Number of clusters: 1363 Maximum cluster size: 4out <- emmeans(mm0,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab$outcome <- "Milk Yield"
tab.my <- tab
tab.my## production contrast estimate outcome
## 1 Low production C_SDCT - BDCT 0.34 (-1.05 to 1.73) Milk Yield
## 2 Low production A_SDCT - BDCT -1.16 (-2.74 to 0.41) Milk Yield
## 3 Mid production C_SDCT - BDCT 0.26 (-1.01 to 1.53) Milk Yield
## 4 Mid production A_SDCT - BDCT -0.32 (-1.68 to 1.04) Milk Yield
## 5 High production C_SDCT - BDCT -1.17 (-2.64 to 0.29) Milk Yield
## 6 High production A_SDCT - BDCT -1.48 (-3.08 to 0.12) Milk Yieldatx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
#emm <- emmeans(lmm, ~Tx*my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm <- emm %>% subset(select=c(my3,Tx,TestDIMcat20,estimate,conf.low,conf.high))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
emm$Tx <- fct_relevel(emm$Tx, "BDCT",
"C_SDCT",
"A_SDCT"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(30,55))) + aes(x=as.numeric(TestDIMcat20), y=estimate, group=Tx, colour=Tx) +
labs(y = "Milk Yield)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$conf.low, ymax=emm$conf.high,colour=Tx,fill=Tx), linetype=0, alpha=0.1) + geom_line(aes(colour=Tx,linetype=Tx),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1") + facet_grid( ~ my3)
curve
Interaction model using linear mixed model
lmm <- lmer(MY ~ Tx*my3 + TestDIMcat20 + DOSCC + Parity + (1|FARMID/CowID),
data=dhi)
summary(lmm)## Linear mixed model fit by REML ['lmerMod']
## Formula: MY ~ Tx * my3 + TestDIMcat20 + DOSCC + Parity + (1 | FARMID/CowID)
## Data: dhi
##
## REML criterion at convergence: 31699
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.522 -0.463 0.019 0.511 4.150
##
## Random effects:
## Groups Name Variance Std.Dev.
## CowID:FARMID (Intercept) 28.7 5.36
## FARMID (Intercept) 12.7 3.56
## Residual 43.8 6.62
## Number of obs: 4551, groups: CowID:FARMID, 1363; FARMID, 16
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 33.6274 1.3459 24.98
## TxA_SDCT -1.3967 0.7923 -1.76
## TxBDCT -0.2261 0.7148 -0.32
## my3Mid production 0.8822 0.7428 1.19
## my3High production 2.9199 0.7343 3.98
## TestDIMcat2030 11.1944 0.3721 30.08
## TestDIMcat2050 12.2344 0.3562 34.35
## TestDIMcat2070 11.8726 0.3535 33.59
## TestDIMcat2090 10.2506 0.3784 27.09
## TestDIMcat20110 7.8865 0.3537 22.30
## DOSCC 0.0388 0.1716 0.23
## Parity2 2.9943 0.4409 6.79
## Parity3 3.2631 0.4603 7.09
## TxA_SDCT:my3Mid production 0.8630 1.0809 0.80
## TxBDCT:my3Mid production -0.1086 1.0173 -0.11
## TxA_SDCT:my3High production 1.1529 1.1018 1.05
## TxBDCT:my3High production 1.3083 1.0182 1.28car::Anova(lmm)## Analysis of Deviance Table (Type II Wald chisquare tests)
##
## Response: MY
## Chisq Df Pr(>Chisq)
## Tx 3.93 2 0.14
## my3 72.05 2 2.3e-16 ***
## TestDIMcat20 1653.27 5 < 2e-16 ***
## DOSCC 0.05 1 0.82
## Parity 66.97 2 2.9e-15 ***
## Tx:my3 3.30 4 0.51
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1out <- emmeans(lmm,pairwise ~ Tx | my3,adjust = "none") %>% confint
tab <- data.frame(
production = out[["contrasts"]][["my3"]],
contrast = out[["contrasts"]][["contrast"]],
est = out[["contrasts"]][["estimate"]],
lcl = out[["contrasts"]][["asymp.LCL"]],
ucl = out[["contrasts"]][["asymp.UCL"]]
)
tab <- tab %>% filter(contrast != "C_SDCT - A_SDCT")
tab <- tab %>% mutate(
estimate = paste0(round(est,2),
" (",
round(lcl,2),
" to ",
round(ucl,2),
")"),
est = NULL,
lcl = NULL,
ucl = NULL
)
tab## production contrast estimate
## 1 Low production C_SDCT - BDCT 0.23 (-1.17 to 1.63)
## 2 Low production A_SDCT - BDCT -1.17 (-2.65 to 0.31)
## 3 Mid production C_SDCT - BDCT 0.33 (-1.1 to 1.77)
## 4 Mid production A_SDCT - BDCT -0.2 (-1.69 to 1.29)
## 5 High production C_SDCT - BDCT -1.08 (-2.5 to 0.33)
## 6 High production A_SDCT - BDCT -1.33 (-2.91 to 0.26)fulltab <- rbind(tab.cm,
tab.cull,
tab.imipc,
tab.imicure,
tab.newimi,
tab.scc,
tab.my)
full.main <- rbind(cm.main,
cull.main,
imipc.main,
imicure.main,
iminew.main,
scc.main,
my.main)
full.main$contrast <- full.main$term
full.main$term <- NULL
tableout <- rbind(fulltab,full.main) %>% arrange(outcome,production)
write.csv(tableout,'tableout2.csv')Milk yield at DO —
IMI at 1-13 DIM
Need to choose covariables
ggplot(qtr, aes(x=DOMY, y=imipc)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
IMI in early lactation
gee <- geeglm(imipc ~ my3, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = imipc ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.22522 0.04249 28.10 1.2e-07 ***
## my3Mid production -0.00908 0.01324 0.47 0.49
## my3High production -0.00688 0.02235 0.09 0.76
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.163
## Number of clusters: 16 Maximum cluster size: 1270gee <- geeglm(imipc ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = imipc ~ my3 + factor(Parity) + DOSCC + Tx, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(imipc) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.09729 0.04462 4.75 0.0292 *
## my3Mid production 0.00473 0.01423 0.11 0.7394
## my3High production 0.01632 0.02317 0.50 0.4811
## factor(Parity)2 0.00897 0.02153 0.17 0.6770
## factor(Parity)3 0.01476 0.01557 0.90 0.3433
## DOSCC 0.02687 0.00858 9.80 0.0017 **
## TxA_SDCT -0.00463 0.01370 0.11 0.7353
## TxBDCT -0.02333 0.02058 1.28 0.2571
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.167
## Number of clusters: 16 Maximum cluster size: 1270tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.0973 0.0446 4.75 0.0292 0.00984 0.185
## 2 my3Mid production 0.00473 0.0142 0.111 0.739 -0.0232 0.0326
## 3 my3High production 0.0163 0.0232 0.496 0.481 -0.0291 0.0617
## 4 factor(Parity)2 0.00897 0.0215 0.174 0.677 -0.0332 0.0512
## 5 factor(Parity)3 0.0148 0.0156 0.898 0.343 -0.0158 0.0453
## 6 DOSCC 0.0269 0.00858 9.80 0.00174 0.0100 0.0437
## 7 TxA_SDCT -0.00463 0.0137 0.114 0.735 -0.0315 0.0222
## 8 TxBDCT -0.0233 0.0206 1.28 0.257 -0.0637 0.0170IMI cure
ggplot(qtr, aes(x=DOMY, y=cure)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
#Univariable
gee <- geeglm(cure ~ my3, data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.86822 0.03074 797.91 <2e-16 ***
## my3Mid production 0.02719 0.01490 3.33 0.068 .
## my3High production 0.00863 0.02560 0.11 0.736
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.306
## Number of clusters: 16 Maximum cluster size: 432#Multivariable
gee <- geeglm(cure ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = cure ~ my3 + factor(Parity) + DOSCC + Tx, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(cure) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.94419 0.05834 261.92 <2e-16 ***
## my3Mid production 0.01655 0.01531 1.17 0.280
## my3High production -0.00132 0.02410 0.00 0.956
## factor(Parity)2 -0.03250 0.01481 4.82 0.028 *
## factor(Parity)3 -0.02221 0.02732 0.66 0.416
## DOSCC -0.01085 0.00714 2.31 0.129
## TxA_SDCT 0.00109 0.01297 0.01 0.933
## TxBDCT -0.00855 0.01849 0.21 0.644
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.35
## Number of clusters: 16 Maximum cluster size: 432tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.944 0.0583 262. 0 0.830 1.06
## 2 my3Mid production 0.0165 0.0153 1.17 0.280 -0.0135 0.0466
## 3 my3High production -0.00132 0.0241 0.00299 0.956 -0.0486 0.0459
## 4 factor(Parity)2 -0.0325 0.0148 4.82 0.0282 -0.0615 -0.00348
## 5 factor(Parity)3 -0.0222 0.0273 0.661 0.416 -0.0758 0.0313
## 6 DOSCC -0.0109 0.00714 2.31 0.129 -0.0249 0.00315
## 7 TxA_SDCT 0.00109 0.0130 0.00699 0.933 -0.0243 0.0265
## 8 TxBDCT -0.00855 0.0185 0.214 0.644 -0.0448 0.0277NewIMI
ggplot(qtr, aes(x=DOMY, y=newimi)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", colour="blue", size=1.5)+
xlab("Milk yield")+
ylab("Probability of IMI at calving")+
theme_bw()
gee <- geeglm(newimi ~ my3, data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ my3, family = binomial(link = "identity"),
## data = qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)),
## id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.2162 0.0462 21.93 2.8e-06 ***
## my3Mid production -0.0103 0.0140 0.54 0.46
## my3High production -0.0135 0.0234 0.33 0.56
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.196
## Number of clusters: 16 Maximum cluster size: 1223gee <- geeglm(newimi ~ my3 + factor(Parity) + DOSCC + Tx, data=qtr %>% filter(!is.na(newimi) & !is.na(Parity) & !is.na(DOSCC)), family=binomial(link="identity"), id=FARMID)
summary(gee)##
## Call:
## geeglm(formula = newimi ~ my3 + factor(Parity) + DOSCC + Tx,
## family = binomial(link = "identity"), data = qtr %>% filter(!is.na(newimi) &
## !is.na(Parity) & !is.na(DOSCC)), id = FARMID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 0.08938 0.04263 4.40 0.0360 *
## my3Mid production 0.00204 0.01516 0.02 0.8929
## my3High production 0.00951 0.02638 0.13 0.7186
## factor(Parity)2 0.01004 0.02256 0.20 0.6563
## factor(Parity)3 0.01270 0.01574 0.65 0.4197
## DOSCC 0.02694 0.00724 13.84 0.0002 ***
## TxA_SDCT -0.00512 0.01656 0.10 0.7573
## TxBDCT -0.02603 0.02135 1.49 0.2226
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1 0.2
## Number of clusters: 16 Maximum cluster size: 1223tidy(gee,conf.int = T)## # A tibble: 8 × 7
## term estimate std.error statistic p.value conf.low conf.high
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 (Intercept) 0.0894 0.0426 4.40 0.0360 0.00583 0.173
## 2 my3Mid production 0.00204 0.0152 0.0181 0.893 -0.0277 0.0318
## 3 my3High production 0.00951 0.0264 0.130 0.719 -0.0422 0.0612
## 4 factor(Parity)2 0.0100 0.0226 0.198 0.656 -0.0342 0.0543
## 5 factor(Parity)3 0.0127 0.0157 0.651 0.420 -0.0181 0.0436
## 6 DOSCC 0.0269 0.00724 13.8 0.000199 0.0127 0.0411
## 7 TxA_SDCT -0.00512 0.0166 0.0955 0.757 -0.0376 0.0273
## 8 TxBDCT -0.0260 0.0213 1.49 0.223 -0.0679 0.0158Post calving SCC
desc <- dhi %>% group_by(my3,TestDIMcat20) %>% summarise(
atrisk = sum(!is.na(scc)),
scm.n = sum(scc > log(200)),
scm.risk = scm.n/atrisk
)
ggplot(desc, aes(x=TestDIMcat20 %>% as.character() %>% as.numeric, y=scm.risk, color = my3)) + geom_point(size=2, alpha=0.4)+
stat_smooth(method="loess", size=1.5)+
xlab("Days in milk")+
ylab("Risk of subclinical mastitis")+
theme_bw()
mm0 <- geeglm(scc ~ my3 + TestDIMcat20 + factor(FARMID), data=dhi, id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(FARMID), data = dhi,
## id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 4.3440 0.1360 1020.15 < 2e-16 ***
## my3Mid production -0.1230 0.0668 3.39 0.0656 .
## my3High production -0.1373 0.0673 4.16 0.0415 *
## TestDIMcat2030 -0.5182 0.0656 62.36 2.9e-15 ***
## TestDIMcat2050 -0.5566 0.0594 87.92 < 2e-16 ***
## TestDIMcat2070 -0.4089 0.0613 44.50 2.5e-11 ***
## TestDIMcat2090 -0.2790 0.0661 17.81 2.4e-05 ***
## TestDIMcat20110 -0.1188 0.0613 3.76 0.0526 .
## factor(FARMID)2 -0.0236 0.1734 0.02 0.8919
## factor(FARMID)3 -0.1044 0.1350 0.60 0.4393
## factor(FARMID)4 0.0372 0.1316 0.08 0.7776
## factor(FARMID)5 0.1338 0.1394 0.92 0.3370
## factor(FARMID)6 -0.5420 0.1784 9.24 0.0024 **
## factor(FARMID)7 -0.3900 0.1507 6.70 0.0097 **
## factor(FARMID)8 0.0359 0.2954 0.01 0.9034
## factor(FARMID)9 -0.1008 0.2962 0.12 0.7337
## factor(FARMID)10 -0.3590 0.2532 2.01 0.1562
## factor(FARMID)11 -0.3386 0.2486 1.86 0.1732
## factor(FARMID)12 -0.5843 0.2279 6.57 0.0104 *
## factor(FARMID)13 0.1225 0.3199 0.15 0.7019
## factor(FARMID)14 0.0831 0.2633 0.10 0.7523
## factor(FARMID)15 -0.0348 0.3940 0.01 0.9296
## factor(FARMID)16 -0.1811 0.2367 0.59 0.4442
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.78 0.0595
## Number of clusters: 1364 Maximum cluster size: 4mm0 <- geeglm(scc ~ my3 + TestDIMcat20 + factor(Parity) + DOSCC + Tx + factor(FARMID), data=dhi, id=CowID)
summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(Parity) +
## DOSCC + Tx + factor(FARMID), data = dhi, id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.351910 0.188167 317.32 < 2e-16 ***
## my3Mid production 0.003227 0.068036 0.00 0.9622
## my3High production 0.039849 0.069379 0.33 0.5657
## TestDIMcat2030 -0.541797 0.064641 70.25 < 2e-16 ***
## TestDIMcat2050 -0.556459 0.059010 88.92 < 2e-16 ***
## TestDIMcat2070 -0.418450 0.060647 47.61 5.2e-12 ***
## TestDIMcat2090 -0.286936 0.064794 19.61 9.5e-06 ***
## TestDIMcat20110 -0.123939 0.061047 4.12 0.0423 *
## factor(Parity)2 0.132579 0.062601 4.49 0.0342 *
## factor(Parity)3 0.302455 0.073465 16.95 3.8e-05 ***
## DOSCC 0.173249 0.026961 41.29 1.3e-10 ***
## TxA_SDCT -0.000951 0.068521 0.00 0.9889
## TxBDCT -0.028799 0.063997 0.20 0.6527
## factor(FARMID)2 0.072952 0.169321 0.19 0.6666
## factor(FARMID)3 -0.019794 0.132363 0.02 0.8811
## factor(FARMID)4 0.063332 0.128095 0.24 0.6210
## factor(FARMID)5 0.031047 0.135709 0.05 0.8190
## factor(FARMID)6 -0.474796 0.176963 7.20 0.0073 **
## factor(FARMID)7 -0.272308 0.146192 3.47 0.0625 .
## factor(FARMID)8 -0.072545 0.279720 0.07 0.7954
## factor(FARMID)9 -0.066661 0.297679 0.05 0.8228
## factor(FARMID)10 -0.244852 0.224595 1.19 0.2756
## factor(FARMID)11 -0.400079 0.239705 2.79 0.0951 .
## factor(FARMID)12 -0.473582 0.216774 4.77 0.0289 *
## factor(FARMID)13 0.117716 0.302994 0.15 0.6976
## factor(FARMID)14 0.104514 0.255829 0.17 0.6829
## factor(FARMID)15 0.031459 0.370484 0.01 0.9323
## factor(FARMID)16 -0.056807 0.232973 0.06 0.8074
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0567
## Number of clusters: 1364 Maximum cluster size: 4summary(mm0)##
## Call:
## geeglm(formula = scc ~ my3 + TestDIMcat20 + factor(Parity) +
## DOSCC + Tx + factor(FARMID), data = dhi, id = CowID)
##
## Coefficients:
## Estimate Std.err Wald Pr(>|W|)
## (Intercept) 3.351910 0.188167 317.32 < 2e-16 ***
## my3Mid production 0.003227 0.068036 0.00 0.9622
## my3High production 0.039849 0.069379 0.33 0.5657
## TestDIMcat2030 -0.541797 0.064641 70.25 < 2e-16 ***
## TestDIMcat2050 -0.556459 0.059010 88.92 < 2e-16 ***
## TestDIMcat2070 -0.418450 0.060647 47.61 5.2e-12 ***
## TestDIMcat2090 -0.286936 0.064794 19.61 9.5e-06 ***
## TestDIMcat20110 -0.123939 0.061047 4.12 0.0423 *
## factor(Parity)2 0.132579 0.062601 4.49 0.0342 *
## factor(Parity)3 0.302455 0.073465 16.95 3.8e-05 ***
## DOSCC 0.173249 0.026961 41.29 1.3e-10 ***
## TxA_SDCT -0.000951 0.068521 0.00 0.9889
## TxBDCT -0.028799 0.063997 0.20 0.6527
## factor(FARMID)2 0.072952 0.169321 0.19 0.6666
## factor(FARMID)3 -0.019794 0.132363 0.02 0.8811
## factor(FARMID)4 0.063332 0.128095 0.24 0.6210
## factor(FARMID)5 0.031047 0.135709 0.05 0.8190
## factor(FARMID)6 -0.474796 0.176963 7.20 0.0073 **
## factor(FARMID)7 -0.272308 0.146192 3.47 0.0625 .
## factor(FARMID)8 -0.072545 0.279720 0.07 0.7954
## factor(FARMID)9 -0.066661 0.297679 0.05 0.8228
## factor(FARMID)10 -0.244852 0.224595 1.19 0.2756
## factor(FARMID)11 -0.400079 0.239705 2.79 0.0951 .
## factor(FARMID)12 -0.473582 0.216774 4.77 0.0289 *
## factor(FARMID)13 0.117716 0.302994 0.15 0.6976
## factor(FARMID)14 0.104514 0.255829 0.17 0.6829
## factor(FARMID)15 0.031459 0.370484 0.01 0.9323
## factor(FARMID)16 -0.056807 0.232973 0.06 0.8074
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation structure = independence
## Estimated Scale Parameters:
##
## Estimate Std.err
## (Intercept) 1.71 0.0567
## Number of clusters: 1364 Maximum cluster size: 4atx <- c(10,30,50,70,90,110)
emm <- emmeans(mm0, ~ my3*TestDIMcat20,adjust = "none", at=list(atx)) %>%confint %>% tidy()
emm$SCC <- exp(emm$estimate)
emm$LCL <- exp(emm$conf.low)
emm$UCL <- exp(emm$conf.high)
emm <- emm %>% subset(select=c(my3,TestDIMcat20,SCC,LCL,UCL))
emm$my3 <- fct_relevel(emm$my3, "Low production",
"Mid production",
"High production"
)
curve <- ggplot(emm) + coord_cartesian(ylim = (c(0,100))) + aes(x=as.numeric(TestDIMcat20), y=SCC, group=my3, colour=my3) +
labs(y = "Somatic cell count \n(geometric mean x 1,000 cells/ml)", x = "Days in milk") + geom_point(size=3) + geom_ribbon(aes(ymin=emm$LCL, ymax=emm$UCL,colour=my3,fill=my3), linetype=0, alpha=0.1) + geom_line(aes(colour=my3,linetype=my3),size=1.1) +
scale_x_continuous(breaks = seq(10, 110, 20)) +
theme(panel.border = element_rect(colour = "black", fill=NA, size=1),axis.text=element_text(size=12,family="Times"),axis.title=element_text(size=14,face="bold",family="Times"),panel.background = element_rect(fill = "white", colour = "white",size = 0.5, linetype = "solid"),panel.grid.major = element_line(size = 0.5, linetype = 'solid', colour = "grey"), panel.grid.minor = element_line(size = 0.25, linetype = 'solid',colour = "grey"),legend.position="bottom",legend.text = element_text(colour="black", size=12,face="bold",family="Times")) + scale_color_brewer(palette="Set1")
curve