Michigan Deer Antler Diameter
Spatio-temporal modeling of areal data with space-time interaction
Gary Roloff
Sean Sultaire
John Humphreys
Contact: John (humph173@msu.edu)
date() ## [1] "Mon Aug 26 15:45:00 2019"Options
library(knitr)
library(rgl)
opts_knit$set(verbose = FALSE)
knit_hooks$set(webgl = hook_webgl)Set Directory
setwd("~/Michigan_State/SLP_Beam_Diam")Load Packages:
suppressMessages(library(INLA))
suppressMessages(library(rgdal))
suppressMessages(library(reshape2))
suppressMessages(library(raster))
suppressMessages(library(fields))
suppressMessages(library(FactoMineR))
suppressMessages(library(missMDA))
suppressMessages(library(rnoaa))
suppressMessages(library(climdex.pcic))
suppressMessages(library(seas))
suppressMessages(library(factoextra))
suppressMessages(library(corrplot))
suppressMessages(library(rasterVis))
suppressMessages(library(rgeos))
suppressMessages(library(maptools))
suppressMessages(library(mapproj))
suppressMessages(library(spdep))
suppressMessages(library(ggplot2))
suppressMessages(library(ggthemes))
suppressMessages(library(GISTools))
suppressMessages(library(lattice))
suppressMessages(library(gridExtra))
suppressMessages(library(xtable))
suppressMessages(library(Rmisc))
suppressMessages(library(dplyr))
suppressMessages(library(dismo))
suppressMessages(library(tseries))
suppressMessages(library(data.table))
suppressMessages(library(readr))
suppressMessages(library(stringr))
suppressMessages(library(kableExtra))
suppressMessages(library(lettercase))
suppressMessages(library(lubridate))
suppressMessages(library(astsa))
source("RScripts.R") Load and Clean Antler Data (1987-2013)
Beamdata = read.csv("./beam_records_all_051118.csv", stringsAsFactors=FALSE) #1987-2013
SexR13 = Beamdata %>%
filter(cseason == "Early Firearm" |
cseason == "Regular Firearm" |
cseason == "Late Firearm") %>%
select(county, year, sex, age)
SexR13$sex = ifelse(SexR13$sex == 1, "M", "F")
MyBeam13 = Beamdata %>%
mutate(LBeam = lbeam,
RBeam = rbeam,
Points = points,
Date = dmy(date)) %>%
select(county, year, Date, LBeam, RBeam, Points, season, sex, age) %>%
filter(county != 99,
age == 1) #selecting yearlingTable to ID Regions
Regs.df = as.data.frame(
Beamdata %>%
group_by(county) %>%
summarize(Reg = median(region)))[1:83,]Load and Clean Antler Data (2014-2016)
Beamdata16 = read.csv("./deeCheckStationData2014_2016.csv", stringsAsFactors=FALSE) #2014-2016
DupCriteria = c("Harvest_Year", "Harvest_Date", "Harvest_Section", "Harvest_TRS", "Gender",
"Age_Abbr", "LBeam_Diameter_Number", "RBeam_Diameter_Number",
"Total_Points_Number", "Data_Collector", "Station_Number")
Beamdata16$Duplicate = duplicated(Beamdata16[, DupCriteria]) | duplicated(Beamdata16[, DupCriteria], fromLast = TRUE)
length(which(Beamdata16$Duplicate == TRUE))## [1] 2503Beamdata16$Dups = duplicated(Beamdata16[, DupCriteria])
SexR16 = Beamdata16 %>%
filter(Dups == FALSE,
Harvest_Season == "Early Firearm " |
Harvest_Season == "Regular Firearm " |
Harvest_Season == "Late Firearm ") %>%
mutate(sex = Gender,
age = Age_Abbr,
year = Harvest_Year,
County = trimws(Harvest_County)) %>%
select(County, year, sex, age)
MyBeam16 = Beamdata16 %>%
mutate(Date = as.Date(Harvest_Date, "%m/%d/%Y"),
County = trimws(Harvest_County),
Points_Number = Total_Points_Number,
year = Harvest_Year,
season = Harvest_Season,
sex = Gender,
age = Age_Abbr) %>%
select(County, year, season, Date, sex, age, LBeam_Diameter, RBeam_Diameter, Dups, Points_Number) %>%
filter(age == 1,
Dups == FALSE,
LBeam_Diameter != "B ",
LBeam_Diameter != "U ",
LBeam_Diameter != "",
is.na(LBeam_Diameter)==FALSE,
RBeam_Diameter != "B ",
RBeam_Diameter != "U ",
RBeam_Diameter != "",
is.na(RBeam_Diameter)==FALSE)
MyBeam16 = MyBeam16 %>%
mutate(LBeam = as.numeric(LBeam_Diameter),
RBeam = as.numeric(RBeam_Diameter),
Points = Points_Number) %>%
select(-c(Points_Number, LBeam_Diameter, RBeam_Diameter, Dups))Load and Clean Antler Data (2017)
Beamdata17 = read.csv("./deeCheckStationData2017.csv", stringsAsFactors=FALSE) #2017 #2014-2016
Beamdata17$Entered_DateTime = as.POSIXct(Beamdata17$Entered_DateTime, format = "%d/%m/%Y %I:%M:%S %p")
DupCriteria = c("Season_Year", "Harvest_Date", "Corrected_Section_Taken", "Beam1_Number", "Beam1_Number",
"Recorder_Name", "Registration_Sex", "Points_Number", "Corrected_Unit", "Station_Number", "Entered_DateTime")
Beamdata17$Duplicate = duplicated(Beamdata17[, DupCriteria]) | duplicated(Beamdata17[, DupCriteria], fromLast = TRUE)
length(which(Beamdata17$Duplicate == TRUE))## [1] 10985Beamdata17$Dups = duplicated(Beamdata17[, DupCriteria])
SexR17 = Beamdata17 %>%
filter(Dups == FALSE,
Harvest_Season == "Early Firearm" |
Harvest_Season == "Regular Firearm" |
Harvest_Season == "Late Firearm") %>%
mutate(sex = Registration_Sex,
age = Registration_Age,
year = Season_Year,
county = Calculated_County_Number_Taken) %>%
select(county, year, sex, age)
MyBeam17 = Beamdata17 %>%
mutate(Date = as.Date(Harvest_Date, "%m/%d/%Y"),
County = trimws(Corrected_County_Name_Taken),
County = gsub("Saint", "St.", County),
year = Season_Year,
season = Harvest_Season,
sex = Registration_Sex,
age = Registration_Age) %>%
select(County, year, season, Date, sex, age, Beam1_Number, Beam2_Number, Dups, Points_Number) %>%
filter(age == 1,
Dups == FALSE,
is.na(Beam1_Number)==FALSE,
is.na(Beam2_Number)==FALSE)
MyBeam17 = MyBeam17 %>%
mutate(LBeam = as.numeric(Beam1_Number),
RBeam = as.numeric(Beam2_Number),
Points = Points_Number) %>%
select(-c(Beam1_Number, Beam2_Number, Points_Number, Dups))Get County Boundaries
Source: http://gis-michigan.opendata.arcgis.com/datasets/counties-v17a
Counties = readOGR(dsn = "./Counties_v17a",
layer = "Counties_v17a",
stringsAsFactors = FALSE)## OGR data source with driver: ESRI Shapefile
## Source: "C:\Users\humph173\Documents\Michigan_State\SLP_Beam_Diam\Counties_v17a", layer: "Counties_v17a"
## with 83 features
## It has 15 fields
## Integer64 fields read as strings: OBJECTID FIPSNUMAdd County Names and Codes
LookUp.tab = Counties@data[, c("OBJECTID", "NAME")]
MyBeam13$County = with(LookUp.tab,
NAME[match(
MyBeam13$county,
OBJECTID)])
MyBeam16$county = with(LookUp.tab,
OBJECTID[match(
MyBeam16$County,
NAME)])
MyBeam17$county = with(LookUp.tab,
OBJECTID[match(
MyBeam17$County,
NAME)])Combine All Years
MyBeam = rbind(MyBeam13, MyBeam16, MyBeam17)Sex Ratio
SexR16$county = with(LookUp.tab,
OBJECTID[match(
SexR16$County,
NAME)])
SexR16 = SexR16 %>% select(-County)
SR.df = rbind(SexR13, SexR16, SexR17) %>%
mutate(county = as.integer(county)) %>%
filter(county >=1 & county <= 83)
SR.df$Age = ifelse(SR.df$age == "0", "Fawn",
ifelse(SR.df$age == "1", "Yearling", "Adult"))
Fawn1 = as.data.frame(
SR.df %>%
filter(Age == "Fawn") %>%
group_by(county, year) %>%
summarize(Fwn = length(age)))
YB1 = as.data.frame(
SR.df %>%
filter(Age == "Yearling",
sex == "M") %>%
group_by(county, year) %>%
summarize(YB = length(age)))
YD1 = as.data.frame(
SR.df %>%
filter(Age == "Yearling",
sex == "F") %>%
group_by(county, year) %>%
summarize(YD = length(age)))
B1 = as.data.frame(
SR.df %>%
filter(Age == "Adult",
sex == "M") %>%
group_by(county, year) %>%
summarize(B = length(age)))
D1 = as.data.frame(
SR.df %>%
filter(Age == "Adult",
sex == "F") %>%
group_by(county, year) %>%
summarize(D = length(age)))
Yr.lvl = levels(factor(SR.df$year))
SR_87_13.df = as.data.frame(
cbind(
county = rep(1:83, each = length(Yr.lvl)),
year = rep(levels(factor(SR.df$year)), n = 1:83)))
SR_87_13.df = join(SR_87_13.df, Fawn1, by=c("county", "year"), type="left")
SR_87_13.df = join(SR_87_13.df, YB1, by=c("county", "year"), type="left")
SR_87_13.df = join(SR_87_13.df, YD1, by=c("county", "year"), type="left")
SR_87_13.df = join(SR_87_13.df, B1, by=c("county", "year"), type="left")
SR_87_13.df = join(SR_87_13.df, D1, by=c("county", "year"), type="left")
SR_87_13.df = SR_87_13.df %>%
mutate(N = B + D,
SR = B/D,
pYB = YB/(YB + B),
pYD = YD/(YD + D),
FD.ratio = Fwn/D,
DB.ratio = pYD/pYB,
B.mort = pYB*0.9,
Avl.B = (B+YB)/0.9,
Avl.D = Avl.B*DB.ratio,
Avl.F = Avl.D*FD.ratio,
N2 = Avl.B + Avl.D + Avl.F)View Counties
Region outlined in red show the 12 counties subject to APRs.
MapCent = gCentroid(Counties, byid=T)
cnt.df = as.data.frame(MapCent@coords)
cnt.df$id = rownames(cnt.df)
cnt.df$Name = Counties$NAME
APR.cnty = gUnaryUnion(
subset(Counties,
NAME == "Antrim" |
NAME == "Benzie" |
NAME == "Charlevoix" |
NAME == "Emmet" |
NAME == "Grand Traverse" |
NAME == "Kalkaska" |
NAME == "Lake" |
NAME == "Manistee" |
NAME == "Mason" |
NAME == "Missaukee" |
NAME == "Osceola" |
NAME == "Wexford"))
wmap_df = fortify(Counties)## Regions defined for each Polygonsapr.cnty = fortify(APR.cnty)
ggplot(wmap_df, aes(long,lat, group=group)) +
geom_polygon(fill = "tan", col="gray21") +
geom_polygon(data=apr.cnty,
aes(long,lat, group=group),
fill = "gray", col="red", alpha=0.5) +
geom_point(data=cnt.df,
aes(x, y,
group=cnt.df$id), size = 0) +
geom_text(data=cnt.df, aes(label = cnt.df$Name,
x = x, y = y,
group=cnt.df$Name),
cex=2, col = "black") +
xlab("Longitude") +
ylab("Latitude") +
coord_equal() +
theme(panel.grid.minor = element_blank(),
panel.grid.major = element_blank(),
panel.background = element_blank(),
plot.background = element_rect(fill = "gray80", linetype="solid"),
panel.border = element_blank(),
legend.title = element_text(size=16, face="bold"),
legend.key = element_rect(fill = "gray80", linetype=0),
legend.background = element_rect(fill = "gray80", linetype=0),
axis.title.x = element_text(size=18, face="bold"),
axis.title.y = element_text(size=18, face="bold"),
axis.text.x = element_text(size=10, face="bold"),
axis.text.y = element_text(size=10, face="bold"),
plot.title = element_blank())
Quick peek at response variable
hist(MyBeam$LBeam)
hist(MyBeam$RBeam)
hist(MyBeam$Points)
Check Data
dim(MyBeam)## [1] 423888 10head(MyBeam)## county year Date LBeam RBeam Points season sex age County
## 1 4 2012 2012-11-29 NA NA NA 1 2 1 Alpena
## 2 4 2012 2012-11-29 NA NA NA 1 2 1 Alpena
## 3 4 2012 2012-11-29 6 5 2 1 1 1 Alpena
## 4 4 2012 2012-11-29 NA NA NA 1 2 1 Alpena
## 5 60 2012 2012-11-29 NA NA NA 1 2 1 Montmorency
## 6 4 2012 2012-11-29 15 15 4 1 1 1 AlpenaCo.names = levels(factor(MyBeam$County))
Co.names## [1] "Alcona" "Alger" "Allegan" "Alpena"
## [5] "Antrim" "Arenac" "Baraga" "Barry"
## [9] "Bay" "Benzie" "Berrien" "Branch"
## [13] "Calhoun" "Cass" "Charlevoix" "Cheboygan"
## [17] "Chippewa" "Clare" "Clinton" "Crawford"
## [21] "Delta" "Dickinson" "Eaton" "Emmet"
## [25] "Genesee" "Gladwin" "Gogebic" "Grand Traverse"
## [29] "Gratiot" "Hillsdale" "Houghton" "Huron"
## [33] "Ingham" "Ionia" "Iosco" "Iron"
## [37] "Isabella" "Jackson" "Kalamazoo" "Kalkaska"
## [41] "Kent" "Keweenaw" "Lake" "Lapeer"
## [45] "Leelanau" "Lenawee" "Livingston" "Luce"
## [49] "Mackinac" "Macomb" "Manistee" "Marquette"
## [53] "Mason" "Mecosta" "Menominee" "Midland"
## [57] "Missaukee" "Monroe" "Montcalm" "Montmorency"
## [61] "Muskegon" "Newaygo" "Oakland" "Oceana"
## [65] "Ogemaw" "Ontonagon" "Osceola" "Oscoda"
## [69] "Otsego" "Ottawa" "Presque Isle" "Roscommon"
## [73] "Saginaw" "Sanilac" "Schoolcraft" "Shiawassee"
## [77] "St. Clair" "St. Joseph" "Tuscola" "Van Buren"
## [81] "Washtenaw" "Wayne" "Wexford"Years.s = levels(factor(MyBeam$year))
Years.s## [1] "1987" "1988" "1989" "1990" "1991" "1992" "1993" "1994" "1995" "1996"
## [11] "1997" "1998" "1999" "2000" "2001" "2002" "2003" "2004" "2005" "2006"
## [21] "2007" "2008" "2009" "2010" "2011" "2012" "2013" "2014" "2015" "2016"
## [31] "2017"Load Crop Data
Source: https://www.nass.usda.gov/Statistics_by_State/Michigan/Publications/County_Estimates/index.php
Crop.df = read.csv("./MI_Crops_1987_2017.csv", stringsAsFactors=FALSE)
Crop.df = Crop.df %>%
select(Year, County, Commodity, Data.Item, Value)
head(Crop.df)## Year County Commodity Data.Item Value
## 1 2017 CLARE CORN CORN - ACRES PLANTED 5,900
## 2 2017 CLARE CORN CORN, GRAIN - ACRES HARVESTED 4,000
## 3 2017 CLARE CORN CORN, GRAIN - PRODUCTION, MEASURED IN BU 580,000
## 4 2017 CLARE CORN CORN, GRAIN - YIELD, MEASURED IN BU / ACRE 145
## 5 2017 CLARE OATS OATS - ACRES HARVESTED 580
## 6 2017 CLARE OATS OATS - ACRES PLANTED 600Keep = c("HARVESTED|PRODUCTION|PLANTED|YIELD")
Crop.df$Item = str_extract(Crop.df$Data.Item, Keep)
Crop.df = Crop.df[complete.cases(Crop.df), ]
Crop.df = Crop.df %>%
filter(County != "OTHER (COMBINED) COUNTIES") %>%
mutate(Year = factor(as.character(Year)),
Variable = paste(Commodity, Item, sep="_"),
Value = gsub("[[:punct:]]", "", Value),
County = str_title_case(str_lowercase(County)),
County = gsub("St", "St.", County),
County = factor(as.character(County)),
Variable = factor(as.character(Variable)),
Value = as.numeric(Value)) %>%
select(-c(Commodity, Data.Item, Item))
Crop.df = dcast(Crop.df,
formula = County+Year ~ Variable,
fill = 0, value.var = "Value",
fun.aggregate = sum, drop = FALSE)
Crop.df[Crop.df == 0] = NA
head(Crop.df)## County Year BEANS_HARVESTED BEANS_PLANTED BEANS_PRODUCTION BEANS_YIELD
## 1 Alcona 1987 380 400 3800 2000
## 2 Alcona 1988 200 300 2000 2000
## 3 Alcona 1989 180 200 2200 1220
## 4 Alcona 1990 250 300 4000 1600
## 5 Alcona 1991 NA NA NA NA
## 6 Alcona 1992 420 500 7000 1670
## CORN_HARVESTED CORN_PLANTED CORN_PRODUCTION CORN_YIELD OATS_HARVESTED
## 1 1900 2100 158500 1140 800
## 2 1870 2000 47000 845 900
## 3 2150 2200 64000 549 1100
## 4 1900 2000 78000 671 780
## 5 2000 2000 106500 680 400
## 6 1800 3000 74500 658 330
## OATS_PLANTED OATS_PRODUCTION OATS_YIELD SOYBEANS_HARVESTED
## 1 1000 41300 516 50
## 2 1300 21500 239 50
## 3 1200 65000 591 80
## 4 900 45000 577 NA
## 5 500 17000 425 NA
## 6 400 18000 545 70
## SOYBEANS_PLANTED SOYBEANS_PRODUCTION SOYBEANS_YIELD SUGARBEETS_HARVESTED
## 1 50 1350 27 NA
## 2 50 1100 22 NA
## 3 80 2100 263 NA
## 4 NA NA NA NA
## 5 NA NA NA NA
## 6 100 850 121 NA
## SUGARBEETS_PLANTED SUGARBEETS_PRODUCTION SUGARBEETS_YIELD
## 1 NA NA NA
## 2 NA NA NA
## 3 NA NA NA
## 4 NA NA NA
## 5 NA NA NA
## 6 NA NA NA
## WHEAT_HARVESTED WHEAT_PLANTED WHEAT_PRODUCTION WHEAT_YIELD
## 1 1000 1200 44000 88
## 2 1800 1900 65000 722
## 3 1700 1800 80000 942
## 4 1380 1400 68000 986
## 5 940 1000 36000 766
## 6 1340 1400 54000 806Reduce Crop Data
Crop.imp = as.data.frame(imputePCA(Crop.df[,3:26],
ncp=22, maxiter = 999)$completeObs)
Crop.pca = PCA(Crop.imp,
scale.unit = TRUE,
ncp = 5,
graph = FALSE)
SVD.df = as.data.frame(Crop.pca$svd$U)
names(SVD.df) = paste("SVD", paste(1:5), sep="")
apply(SVD.df,2,range)## SVD1 SVD2 SVD3 SVD4 SVD5
## [1,] -1.665620 -4.179704 -4.569588 -8.145038 -6.353382
## [2,] 7.136421 7.903224 14.350771 4.975552 4.533176Crop.df = cbind(Crop.df, SVD.df)
CorTab = cor(Crop.df[,3:31], use="pairwise.complete")
corrplot(CorTab)
Clean Crop Data
Also adding to antler data.
Yr.lv = levels(factor(Crop.df$Year))
for(i in 1:length(Yr.lv)){
tLU.tab = Crop.df %>%
filter(Year == Yr.lv[i])
tBm.df = MyBeam %>%
filter(year == Yr.lv[i])
kp.df = join(tBm.df, tLU.tab, by="County", type="left")
if(i == 1){MyBeam2 = kp.df}
else{MyBeam2 = rbind(MyBeam2, kp.df)}
}
MyBeam2$Year = MyBeam2$yearScale and Center Crops
Cov.df = MyBeam2[,12:35]
for(i in 1:ncol(Cov.df)){
Cov.df[,i] = as.numeric(scale(Cov.df[,i], scale = T, center = T))
}
names(Cov.df) = paste("s", names(Cov.df), sep="")
apply(Cov.df, 2, range, na.rm=T)## sBEANS_HARVESTED sBEANS_PLANTED sBEANS_PRODUCTION sBEANS_YIELD
## [1,] -0.516943 -0.5266704 -0.4802203 -1.395671
## [2,] 6.631812 6.3816079 9.0315488 3.315625
## sCORN_HARVESTED sCORN_PLANTED sCORN_PRODUCTION sCORN_YIELD
## [1,] -1.015392 -1.032370 -0.9000729 -1.008246
## [2,] 3.701332 3.547101 3.2989534 2.967560
## sOATS_HARVESTED sOATS_PLANTED sOATS_PRODUCTION sOATS_YIELD
## [1,] -0.7240608 -0.7678715 -0.6243832 -0.8351188
## [2,] 15.6707788 14.1562125 19.0740135 3.4519994
## sSOYBEANS_HARVESTED sSOYBEANS_PLANTED sSOYBEANS_PRODUCTION
## [1,] -1.005589 -1.006778 -0.960451
## [2,] 2.979522 2.979297 3.682223
## sSOYBEANS_YIELD sSUGARBEETS_HARVESTED sSUGARBEETS_PLANTED
## [1,] -0.9376649 -0.8963345 -0.9034098
## [2,] 2.2643967 3.2545439 3.2181718
## sSUGARBEETS_PRODUCTION sSUGARBEETS_YIELD sWHEAT_HARVESTED
## [1,] -0.8348587 -1.861187 -0.7790677
## [2,] 4.5159958 1.980259 4.8728372
## sWHEAT_PLANTED sWHEAT_PRODUCTION sWHEAT_YIELD
## [1,] -0.7862516 -0.6990415 -0.9665892
## [2,] 5.0319243 5.6534157 3.0431404MyBeam2 = cbind(MyBeam2, Cov.df)Load and Clean PRISM Monthy Climate Data
First = read.csv("./PRISM_series1_198701_200112.csv", stringsAsFactors=FALSE)
Second = read.csv("./PRISM_series2_200201_201312.csv", stringsAsFactors=FALSE)
Third = read.csv("./PRISM_series3_201401_201712.csv", stringsAsFactors=FALSE)
PRISM = rbind(First, Second, Third)
PRISM = PRISM %>%
mutate(Idx = 1:nrow(PRISM),
Year = as.integer(substr(Date, 1, 4)),
Month = as.integer(substr(Date, 6, 7)))
PRISM$Season = ifelse(PRISM$Month == 12 | PRISM$Month <= 2, "Winter",
ifelse(PRISM$Month >= 3 & PRISM$Month <= 5, "Spring",
ifelse(PRISM$Month >= 6 & PRISM$Month <= 8, "Summer",
ifelse(PRISM$Month >= 9 & PRISM$Month <= 11, "Fall",
NA))))Calculate Climatic Variables.
Cnty_lvs = levels(factor(PRISM$Name))
Yr_lvs = levels(factor(PRISM$Year))
Seas_lvs = levels(factor(PRISM$Season))
for(i in 1:length(Cnty_lvs)){
tmp.df = PRISM %>% filter(Name == Cnty_lvs[i])
for(j in 1:length(Yr_lvs)){
tmp.df2 = tmp.df %>% filter(Year == Yr_lvs[j])
Clim.df = as.data.frame(
tmp.df2 %>%
group_by(Season) %>%
summarise(PPTMax = max(ppt_in),
PPTMin = min(ppt_in),
PPTAvg = mean(ppt_in),
TempMax = max(tmax_f),
TempMin = min(tmin_f),
TempAvg = min(tmean_f),
DewAvg = mean(tdmean_f),
VPMax = max(vpdmax_hPa),
VPMin = min(vpdmin_hPa)))
Clim.F = Clim.df[1,2:10]
names(Clim.F) = paste("F_", names(Clim.df[2:10]), sep="")
Clim.Sp = Clim.df[2,2:10]
names(Clim.Sp) = paste("Sp_", names(Clim.df[2:10]), sep="")
Clim.Su = Clim.df[3,2:10]
names(Clim.Su) = paste("Su_", names(Clim.df[2:10]), sep="")
Clim.W = Clim.df[4,2:10]
names(Clim.W) = paste("W_", names(Clim.df[2:10]), sep="")
BioClim.df = as.data.frame(
biovars(tmp.df2$ppt_in,
tmp.df2$tmin_f,
tmp.df2$tmax_f))
ClimVar = cbind(Clim.F, Clim.Sp, Clim.Su, Clim.W, BioClim.df)
ClimVar$Year2 = Yr_lvs[j]
ClimVar$county2 = Cnty_lvs[i]
if(j == 1){Yr.df = ClimVar}
else{Yr.df = rbind(Yr.df, ClimVar)}
}
if(i == 1){FClim.df = Yr.df}
else{FClim.df = rbind(FClim.df, Yr.df)}
}
FClim.df$County = with(LookUp.tab,
NAME[match(
FClim.df$county2,
OBJECTID)])Scale and Center
Cov.df = FClim.df[,1:55]
for(i in 1:ncol(Cov.df)){
Cov.df[,i] = as.numeric(scale(Cov.df[,i], scale = T, center = T))
}
names(Cov.df) = paste("s", names(Cov.df), sep="")
apply(Cov.df, 2, range, na.rm=T)## sF_PPTMax sF_PPTMin sF_PPTAvg sF_TempMax sF_TempMin sF_TempAvg
## [1,] -2.153105 -2.065611 -2.249690 -3.181243 -4.146686 -3.770235
## [2,] 5.351474 4.667946 4.251536 2.443820 2.379860 2.415509
## sF_DewAvg sF_VPMax sF_VPMin sSp_PPTMax sSp_PPTMin sSp_PPTAvg
## [1,] -3.920180 -2.206683 -1.612481 -2.103558 -1.902878 -2.407476
## [2,] 2.976639 3.299410 5.183839 4.994053 4.047536 3.926692
## sSp_TempMax sSp_TempMin sSp_TempAvg sSp_DewAvg sSp_VPMax sSp_VPMin
## [1,] -3.258150 -4.113935 -3.661339 -3.190592 -2.741134 -2.168546
## [2,] 2.266514 2.823693 3.113358 2.679437 2.578725 5.000602
## sSu_PPTMax sSu_PPTMin sSu_PPTAvg sSu_TempMax sSu_TempMin sSu_TempAvg
## [1,] -2.215521 -2.306875 -2.565763 -3.443568 -2.731916 -2.772590
## [2,] 5.862130 3.949674 3.877409 2.983498 2.682124 2.667825
## sSu_DewAvg sSu_VPMax sSu_VPMin sW_PPTMax sW_PPTMin sW_PPTAvg
## [1,] -4.021431 -2.793244 -1.444566 -2.196807 -1.924136 -2.166591
## [2,] 2.776874 4.118927 5.503902 4.484602 5.221264 5.073001
## sW_TempMax sW_TempMin sW_TempAvg sW_DewAvg sW_VPMax sW_VPMin
## [1,] -3.375214 -3.480738 -3.392307 -3.234305 -2.84207 -2.184457
## [2,] 2.923848 2.424367 2.321205 2.801817 4.70639 3.867811
## sbio1 sbio2 sbio3 sbio4 sbio5 sbio6 sbio7
## [1,] -3.441152 -3.115318 -3.205317 -2.593757 -3.446926 -3.480215 -2.828364
## [2,] 2.416824 3.213514 3.609070 3.593606 2.985402 2.431332 3.515457
## sbio8 sbio9 sbio10 sbio11 sbio12 sbio13 sbio14
## [1,] -3.997229 -1.810934 -3.224679 -3.499509 -2.870396 -2.153374 -2.018767
## [2,] 1.656631 2.941832 2.460256 2.511412 4.128713 5.156541 4.005414
## sbio15 sbio16 sbio17 sbio18 sbio19
## [1,] -2.914072 -2.523214 -2.557151 -2.503395 -2.058034
## [2,] 4.070966 3.963281 3.806903 3.871831 5.040367FClim.df = cbind(FClim.df, Cov.df)Reduce Climate Data
Clims.pca = PCA(FClim.df[,95:113],
scale.unit = TRUE,
ncp = 5,
graph = FALSE)
SVD2.df = as.data.frame(Clims.pca$svd$U)
names(SVD2.df) = paste("SVD2", paste(1:5), sep="")
FClim.df = cbind(FClim.df, SVD2.df)Combine Climate to Antler Data
Yr.lv = levels(factor(FClim.df$Year))
for(i in 1:length(Yr.lv)){
tLU.tab = FClim.df %>%
filter(Year2 == Yr.lv[i])
tBm.df = MyBeam2 %>%
filter(year == Yr.lv[i])
kp.df = join(tBm.df, tLU.tab, by="County", type="left")
if(i == 1){MyBeam3 = kp.df}
else{MyBeam3 = rbind(MyBeam3, kp.df)}
}
MyBeam3 = MyBeam3 %>%
select(-c(county2, Year2))Regional Climate Analysis
Station Locations
Stat.loc = read.csv("./MI_WI_11-6B.csv", stringsAsFactors=FALSE)
Stat.loc = Stat.loc[!duplicated(Stat.loc$STATION), ] %>%
filter(LONGITUDE != "unknown",
LATITUDE != "unknown") %>%
mutate(LONGITUDE = as.numeric(LONGITUDE),
LATITUDE = as.numeric(LATITUDE))
Locs = cbind(Stat.loc$LONGITUDE, Stat.loc$LATITUDE)
Stat.pnts = SpatialPointsDataFrame(Locs , Stat.loc)
proj4string(Stat.pnts) = proj4string(Counties)
plot(Counties)
plot(Stat.pnts, col = "red", add=T)
Get NOAA Historical Climate Data (GHCN)
Previously run, time intensive.
#Stations = substr(unique(read.csv("./MI_WI_11-6B.csv", stringsAsFactors=FALSE)$STATION), 7, 17)
#Clim.df = meteo_pull_monitors(Stations,
# date_min = "1987-01-01",
# date_max = "2017-12-31")
#Clim.df = as.data.frame(Clim.df) %>%
# select(id, date, prcp, snow, snwd, tavg, tmax, tmin, dapr, mdpr, mdsf)
#write.csv(Clim.df, "./NOAA_Clim.csv")
#token "VmLVwVYuKJixCseSiebNvcoObqeMirtI"Climate Time Series Indicies
Snow.df = read.csv("./NOAA_Clim.csv", stringsAsFactors=FALSE) %>% #preprocessed above
mutate(date = as.Date(date, "%Y-%m-%d"),
Date = as.PCICt(as.character(date),
cal = "gregorian",
format = "%Y-%m-%d"),
Year = year(date),
Month = month(date),
Day = day(date),
DOY = yday(date),
precip = prcp/10, #to mm
t_max = tmax/10, #to C
t_min = tmin/10) #to C
Snow.df[is.na(Snow.df)] = 0
Snow.df$t_avg = (Snow.df$t_max + Snow.df$t_min)/2
Yr.lvs = levels(factor(Snow.df$Year))
loc.lvs = levels(factor(unique(Snow.df$id)))
for(i in 1:length(loc.lvs)){
snw2.tmp = Snow.df %>% filter(id == loc.lvs[i])
Srt.yr = as.numeric(min(levels(factor(snw2.tmp$Year))))
End.yr = as.numeric(max(levels(factor(snw2.tmp$Year))))
#ClimDex Extremes
ci = climdexInput.raw(snw2.tmp$t_max,
snw2.tmp$t_min,
snw2.tmp$precip,
snw2.tmp$Date,
snw2.tmp$Date,
snw2.tmp$Date,
base.range=c(Srt.yr, End.yr))
Frm1.df = as.data.frame(
cbind(
GrowSeas = climdex.gsl(ci, gsl.mode = "GSL"), #Grow season length
PrecEx10mm = climdex.r10mm(ci), #precip exceeding 10mm per day
PrecEx20mm = climdex.r20mm(ci), #precip exceeding 20mm per day
DSubZero = climdex.id(ci), #count of days maximum temp below 0
Prec95pct = climdex.r95ptot(ci), #precip over 95th %
MonMxDT = climdex.txx(ci, freq = "annual"), #Annual Max Daily temp
MMx1DPrec = climdex.rx1day(ci, freq = "annual"), #Ann Maximum 1-day Precipitation
AMinT = climdex.tnn(ci, freq = "annual"), #Annual Minimum Of Daily Minimum Temperature
AMxT = climdex.txn(ci, freq = "annual"), #Annual Minimum Of Daily Maximum Temperature
SumDur = climdex.su(ci), #Summer duration by year
FrstDays = climdex.fd(ci))) #Frost days
Frm1.df$Year = row.names(Frm1.df)
if(range(snw2.tmp$precip)[1] == 0 & range(snw2.tmp$precip)[2] == 0){
Frm1.df$PrecDep = NA
Frm1.df$SnwDep = NA}
else{
#Precip Departure
dat.ss = seas.sum(snw2.tmp, var = "precip", width="month")
dat.sn = seas.norm(dat.ss, var = "precip", fun = "median", norm = "days")
dat.dep = precip.dep(snw2.tmp, dat.sn, var="precip" )
PrecDep = as.data.frame(
dat.dep %>%
group_by(Year) %>%
summarize(Dep = mean(dep, na.rm=T)))
Frm1.df$PrecDep = with(PrecDep,
Dep[match(
Frm1.df$Year,
Year)])
#Snow Departure
dat.ss = seas.sum(snw2.tmp, var = "snow", width="month")
dat.sn = seas.norm(dat.ss, var = "snow", fun = "median", norm = "days")
dat.dep = precip.dep(snw2.tmp, dat.sn, var="snow" )
SnwDep = as.data.frame(
dat.dep %>%
group_by(Year) %>%
summarize(Dep = mean(dep, na.rm=T)))
Frm1.df$SnwDep = with(SnwDep,
Dep[match(
Frm1.df$Year,
Year)])
}
Frm1.df$Station = paste(loc.lvs[i])
if(i == 1){ClimChn = Frm1.df}
else{ClimChn = rbind(ClimChn, Frm1.df)}
}Combine Climate Indices to Antler Data
Cnty.pnts = SpatialPointsDataFrame(cnt.df[,1:2], cnt.df)
proj4string(Cnty.pnts) = proj4string(Stat.pnts)
suppressMessages(library(spatstat))
nProj = "+proj=utm +zone=10 +datum=NAD83 +ellps=GRS80 +towgs84=0,0,0"
Cnty.pp = spTransform(Cnty.pnts, nProj)
Stat.pp = spTransform(Stat.pnts, nProj)
Cnty.pp = as.ppp(Cnty.pp)
Cnty.pp$Name = cnt.df$Name
Stat.pp = as.ppp(Stat.pp)
Stat.pp$Name = substr(Stat.pnts$STATION, 7, 17)
Dist.df = as.data.frame(crossdist(Cnty.pp, Stat.pp))
colnames(Dist.df) = Stat.pp$Name
row.names(Dist.df) = Cnty.pp$Name
Cnty.lvs = levels(factor(Cnty.pp$Name))
for(i in 1:length(Yr.lv)){
tLU.tab = ClimChn %>%
filter(Year == Yr.lv[i])
tBm.df = MyBeam3 %>%
filter(year == Yr.lv[i])
for(j in 1:length(Cnty.lvs)){
tBm2.df = tBm.df %>%
filter(County == Cnty.lvs[j])
if(dim(tBm2.df)[1] == 0){next}
else{
prox = arrange(melt(Dist.df %>%
mutate(Cnty = row.names(Dist.df)) %>%
filter(Cnty == Cnty.lvs[j]), "Cnty"), value)
tLU.tab$Wt = with(prox,
value[match(
tLU.tab$Station,
variable)])
tLU.tab = arrange(tLU.tab[complete.cases(tLU.tab),], Wt)[1:4,]
tLU.tab$WtPro = 1-Scale(tLU.tab$Wt)
W.mean = function(x){weighted.mean(x, tLU.tab$WtPro, na.rm = T)}
Vals = apply(tLU.tab[, c(1:11, 13:14)], 2, W.mean)
An.sm = as.data.frame(matrix(nrow = dim(tBm2.df)[1], ncol = 13))
colnames(An.sm) = names(Vals)
for(tx in 1:dim(tBm2.df)[1]){
An.sm[tx,] = as.numeric(apply(tLU.tab[, c(1:11, 13:14)], 2, W.mean))}
tBm2.df = cbind(tBm2.df, An.sm)
if(j == 1){Cnty.df = tBm2.df}
else{Cnty.df = rbind(Cnty.df, tBm2.df)}
}
}
if(i == 1){Yr.frm = Cnty.df}
else{Yr.frm = rbind(Yr.frm, Cnty.df)}
}
MyBeam3 = Yr.frm
Cov.df = Yr.frm[,180:192]
for(i in 1:ncol(Cov.df)){
Cov.df[,i] = as.numeric(scale(Cov.df[,i], scale = T, center = T))
}
names(Cov.df) = paste("s", names(Cov.df), sep="")
apply(Cov.df, 2, range, na.rm=T)
MyBeam3 = cbind(MyBeam3, Cov.df)Add Sex Ratio
SR_87_13X = SR_87_13.df %>%
mutate(year = as.character(year),
county = as.character(county)) %>%
select(county, year, DB.ratio, FD.ratio)
MyBeam3 = join(MyBeam3, SR_87_13X, by=c("county", "year"), type="left")Average Beam Diameter
MyBeam3$AvgBeam = ifelse(is.na(MyBeam3$RBeam)==TRUE & is.na(MyBeam3$LBeam)==TRUE, NA,
ifelse(is.na(MyBeam3$RBeam)==TRUE & is.na(MyBeam3$LBeam)==FALSE, MyBeam3$LBeam,
ifelse(is.na(MyBeam3$RBeam)==FALSE & is.na(MyBeam3$LBeam)==TRUE, MyBeam3$RBeam,
ifelse(is.na(MyBeam3$RBeam)==FALSE & is.na(MyBeam3$LBeam)==FALSE,
(MyBeam3$LBeam + MyBeam3$RBeam)/2, NA))))Oceanic Nino Index
http://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php
ONI.df = read.csv("C:/Users/humph173/Documents/NOAA_ONI/ONI.csv", stringsAsFactors=FALSE) %>%
filter(Year >= 1987 & Year <= 2017)
MyBeam3$Month = month(MyBeam3$Date)
ONI.lookup = as.data.frame(
cbind(names(ONI.df)[2:13], 1:12))
names(ONI.lookup) = c("Seas", "Month")
MyBeam3$ONISeas = with(ONI.lookup,
Seas[match(
MyBeam3$Month,
Month)])
MyBeam3$ONISeas[is.na(MyBeam3$ONISeas)] = "OND"
for(i in 1:length(Yr.lv)){
tmp.frm = MyBeam3 %>% filter(Year == Yr.lv[i])
ONI.tmp = melt(ONI.df %>% filter(Year == Yr.lv[i]), "Year")
tmp.frm$ONI.indx = with(ONI.tmp,
value[match(
tmp.frm$ONISeas,
variable)])
if(i == 1){OutRes = tmp.frm}
else{OutRes = rbind(OutRes, tmp.frm)}
}
MyBeam3 = OutResSpatio-Temporal Model Construction
Temporal Index
MyBeam3$Year1 = MyBeam3$Year2 =
MyBeam3$Year3 =
MyBeam3$Year4 =
MyBeam3$Year5 =
MyBeam3$Year6 =
MyBeam3$Year.int = as.integer(as.factor(MyBeam3$year))Spatial Index
Counties$Region = 1:nrow(Counties@data)
Region.tab = Counties@data[, c("NAME", "Region")]
MyBeam3$Region = MyBeam3$Region1 =
MyBeam3$Region2 =
MyBeam3$Region3 =
MyBeam3$Region4 =
MyBeam3$Region5 =
MyBeam3$Region6 =
MyBeam3$Region.int =
with(Region.tab,
Region[match(
as.character(MyBeam3$County),
NAME)])Space-Time Interaction Index
MyBeam3$Region.Yr = paste(MyBeam3$County, MyBeam3$year, sep="")County Grouping Effect
Grouping Counties Upper, Lower, and NW12 subject to APRs in 2013.
##The NW 12
MyBeam3$APR = ifelse(MyBeam3$County == "Antrim" |
MyBeam3$County == "Benzie" |
MyBeam3$County == "Charlevoix" |
MyBeam3$County == "Emmet" |
MyBeam3$County == "Grand Traverse" |
MyBeam3$County == "Kalkaska" |
MyBeam3$County == "Lake" |
MyBeam3$County == "Manistee" |
MyBeam3$County == "Mason" |
MyBeam3$County == "Missaukee" |
MyBeam3$County == "Osceola" |
MyBeam3$County == "Wexford", "NW12", "Other")
MyBeam3$Pen = with(Counties@data, #Upper vs. lower
PENINSULA[match(
MyBeam3$county,
OBJECTID)])
MyBeam3$RegAPR = ifelse(MyBeam3$APR != "NW12", MyBeam3$Pen, "APR") #Upper, Lowern and NW12Spatial Neighborhood Graph
nb = poly2nb(Counties,
queen = T)
nb2INLA("J", nb)
J = inla.read.graph("J")
summary(nb)## Neighbour list object:
## Number of regions: 83
## Number of nonzero links: 420
## Percentage nonzero weights: 6.096676
## Average number of links: 5.060241
## Link number distribution:
##
## 1 2 3 4 5 6 7 8
## 1 6 8 21 8 17 19 3
## 1 least connected region:
## 41 with 1 link
## 3 most connected regions:
## 19 61 67 with 8 linksView Spatial Neighborhood Graph
plot(Counties, col='gray', border='blue')
xy = coordinates(Counties)
plot(nb, xy, col='red', lwd=2, add=TRUE)
Check for Colinearity
library(perturb)##
## Attaching package: 'perturb'## The following object is masked from 'package:raster':
##
## reclassifyColin.df = MyBeam3 %>%
select(sbio1, sbio2, sW_PPTMax,sCORN_PLANTED, sWHEAT_PLANTED,
sSOYBEANS_PLANTED, sGrowSeas, sPrecEx10mm, sPrecEx20mm, sDSubZero, sPrec95pct, sMonMxDT,
sMMx1DPrec, sAMinT, sAMxT, sSumDur, sFrstDays, sPrecDep, sSnwDep)
Colin.df = Colin.df[complete.cases(Colin.df),]
CorCov = cor(Colin.df)
corrplot(CorCov)
CI = colldiag(CorCov)
CI## Condition
## Index Variance Decomposition Proportions
## intercept sbio1 sbio2 sW_PPTMax sCORN_PLANTED sWHEAT_PLANTED
## 1 1.000 0.001 0.001 0.001 0.001 0.000 0.000
## 2 1.386 0.000 0.000 0.001 0.001 0.000 0.000
## 3 1.537 0.001 0.000 0.001 0.001 0.001 0.001
## 4 1.928 0.006 0.000 0.001 0.000 0.000 0.000
## 5 2.190 0.001 0.000 0.001 0.013 0.000 0.002
## 6 2.507 0.048 0.000 0.001 0.018 0.000 0.001
## 7 3.337 0.078 0.000 0.005 0.024 0.000 0.001
## 8 3.550 0.020 0.000 0.006 0.082 0.000 0.001
## 9 4.539 0.103 0.004 0.016 0.015 0.000 0.002
## 10 4.826 0.005 0.000 0.014 0.000 0.002 0.007
## 11 6.740 0.004 0.030 0.094 0.066 0.000 0.044
## 12 10.735 0.022 0.031 0.318 0.011 0.012 0.015
## 13 11.925 0.030 0.004 0.003 0.152 0.006 0.002
## 14 13.618 0.021 0.127 0.129 0.136 0.004 0.123
## 15 14.575 0.012 0.003 0.002 0.016 0.012 0.037
## 16 18.778 0.237 0.103 0.017 0.193 0.025 0.012
## 17 23.028 0.190 0.019 0.325 0.180 0.518 0.093
## 18 25.098 0.001 0.014 0.024 0.073 0.116 0.007
## 19 29.218 0.222 0.664 0.040 0.020 0.303 0.652
## sSOYBEANS_PLANTED sGrowSeas sPrecEx10mm sPrecEx20mm sDSubZero
## 1 0.000 0.000 0.000 0.000 0.001
## 2 0.000 0.000 0.000 0.000 0.001
## 3 0.001 0.001 0.002 0.001 0.000
## 4 0.000 0.001 0.001 0.000 0.000
## 5 0.000 0.003 0.000 0.000 0.002
## 6 0.000 0.001 0.001 0.001 0.004
## 7 0.000 0.002 0.003 0.000 0.000
## 8 0.000 0.009 0.001 0.000 0.000
## 9 0.000 0.001 0.004 0.000 0.000
## 10 0.000 0.000 0.005 0.012 0.002
## 11 0.000 0.009 0.008 0.001 0.016
## 12 0.038 0.003 0.001 0.015 0.235
## 13 0.001 0.003 0.221 0.008 0.052
## 14 0.052 0.042 0.117 0.030 0.038
## 15 0.159 0.004 0.296 0.034 0.124
## 16 0.275 0.061 0.011 0.106 0.027
## 17 0.143 0.061 0.166 0.142 0.428
## 18 0.241 0.060 0.047 0.469 0.003
## 19 0.087 0.741 0.115 0.180 0.067
## sPrec95pct sMonMxDT sMMx1DPrec sAMinT sAMxT sSumDur sFrstDays sPrecDep
## 1 0.000 0.000 0.000 0.001 0.000 0.000 0.001 0.000
## 2 0.000 0.002 0.001 0.000 0.001 0.001 0.000 0.000
## 3 0.002 0.000 0.001 0.000 0.000 0.000 0.000 0.000
## 4 0.002 0.000 0.010 0.001 0.001 0.001 0.000 0.010
## 5 0.000 0.004 0.000 0.000 0.001 0.001 0.000 0.032
## 6 0.000 0.006 0.000 0.000 0.001 0.002 0.002 0.008
## 7 0.007 0.005 0.005 0.001 0.006 0.001 0.000 0.029
## 8 0.000 0.001 0.001 0.000 0.002 0.000 0.021 0.003
## 9 0.001 0.048 0.012 0.003 0.010 0.003 0.001 0.001
## 10 0.001 0.023 0.008 0.018 0.016 0.001 0.001 0.051
## 11 0.006 0.002 0.041 0.026 0.001 0.003 0.017 0.017
## 12 0.015 0.013 0.005 0.097 0.002 0.000 0.028 0.007
## 13 0.051 0.008 0.083 0.085 0.151 0.035 0.197 0.001
## 14 0.028 0.000 0.010 0.145 0.074 0.061 0.028 0.000
## 15 0.098 0.007 0.166 0.082 0.160 0.001 0.068 0.008
## 16 0.002 0.052 0.175 0.051 0.013 0.546 0.074 0.345
## 17 0.069 0.002 0.120 0.110 0.011 0.045 0.171 0.430
## 18 0.718 0.407 0.156 0.342 0.377 0.161 0.003 0.007
## 19 0.000 0.419 0.206 0.038 0.174 0.137 0.388 0.051
## sSnwDep
## 1 0.000
## 2 0.006
## 3 0.001
## 4 0.007
## 5 0.004
## 6 0.014
## 7 0.009
## 8 0.045
## 9 0.070
## 10 0.105
## 11 0.006
## 12 0.164
## 13 0.056
## 14 0.125
## 15 0.153
## 16 0.021
## 17 0.027
## 18 0.184
## 19 0.003Model Fitting Set
MyBeam3.tst = MyBeam3 %>% filter(is.na(AvgBeam)==FALSE)
MyBeam3.tst$ID.Region.Yr = as.integer(as.factor(MyBeam3.tst$Region.Yr))Setup Matrix and Priors
Sr = length(unique(MyBeam3.tst$Region))
Tm = length(unique(MyBeam3.tst$Year))
g = inla.read.graph("J")
Q.xi = matrix(0, g$n, g$n)
for (i in 1:g$n){
Q.xi[i,i]=g$nnbs[[i]]
Q.xi[i,g$nbs[[i]]]=-1}
Q.Leroux = diag(Sr)-Q.xi
lunif = "expression:
a=1; b=1;
beta = exp(theta)/(1+exp(theta));
logdens = lgamma(a+b)-lgamma(a)-lgamma(b)+(a-1)*beta+(b-1)*(1-beta);
log_jacobian = log(beta*(1-beta));
return(logdens+log_jacobian)"
sdunif = "expression: logdens = -log_precision/2;
return(logdens)"Spatial Model Only
formula.sp = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif)))
#theta1 = Model.sp$internal.summary.hyperpar$mean
theta1 = c(-2.686110, -2.946664, 3.201317)
Model.sp = inla(formula.sp,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m1 = c(Model.sp$dic$dic, Model.sp$waic$waic)
summary(Model.sp)##
## Call:
## c("inla(formula = formula.sp, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 2.0736 481.8719 18.4135 502.3590
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode kld
## (Intercept) 12.0155 0.0064 12.0028 12.0155 12.0281 12.0155 0
##
## Random effects:
## Name Model
## Region Generic1 model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0681 0.00 0.0681 0.0681
## Precision for Region 0.0525 0.00 0.0525 0.0525
## Beta for Region 0.9609 0.00 0.9609 0.9609
## 0.975quant mode
## Precision for the Gaussian observations 0.0682 0.0681
## Precision for Region 0.0526 0.0525
## Beta for Region 0.9609 0.9609
##
## Expected number of effective parameters(std dev): 83.10(0.00)
## Number of equivalent replicates : 4192.22
##
## Deviance Information Criterion (DIC) ...............: 2793711.65
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 56.14
##
## Watanabe-Akaike information criterion (WAIC) ...: 2793999.26
## Effective number of parameters .................: 343.38
##
## Marginal log-Likelihood: -2324285.42
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedPlot Random Field
LU.tab = as.data.frame(Model.sp$summary.random$Region[, c(1:2,4,6)])
names(LU.tab) = c("ID", "Mean", "Q025", "Q975")
LU.tab$County = LookUp.tab$NAME
Counties2 = Counties
Domain.r = raster(res=0.01,
ext = extent(Counties))
Counties2$RE = LU.tab$Mean
Region.mean.r = rasterize(Counties2,
Domain.r,
field = "RE",
background = NA)
Counties2$Q025 = LU.tab$Q025
Region.Q025.r = rasterize(Counties2,
Domain.r,
field = "Q025",
background = NA)
Counties2$Q975 = LU.tab$Q975
Region.Q975.r = rasterize(Counties2,
Domain.r,
field = "Q975",
background = NA)
MRF.stk = stack(Region.Q025.r, Region.mean.r, Region.Q975.r)
names(MRF.stk) = c("A", "B", "C")
MRF.stk2 = MRF.stkSpace and Non-Parametric Time
formula.st = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif)))
#theta1 = Model.st$internal.summary.hyperpar$mean
theta1 = c(-2.421195, -1.292276, 3.282074, 1.992129)
Model.st = inla(formula.st,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m2 = c(Model.st$dic$dic, Model.st$waic$waic)
summary(Model.st)##
## Call:
## c("inla(formula = formula.st, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 2.7041 744.8421 17.4182 764.9644
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode kld
## (Intercept) 19.5178 0.0078 19.5024 19.5178 19.5332 19.5178 0
##
## Random effects:
## Name Model
## Region Generic1 model
## Year1 RW1 model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0888 0.0000 0.0888 0.0888
## Precision for Region 0.2752 0.0177 0.2340 0.2782
## Beta for Region 0.9637 0.0033 0.9534 0.9645
## Precision for Year1 7.3807 0.8267 5.5107 7.5062
## 0.975quant mode
## Precision for the Gaussian observations 0.0900 0.0888
## Precision for Region 0.3004 0.2945
## Beta for Region 0.9686 0.9650
## Precision for Year1 8.5715 8.2319
##
## Expected number of effective parameters(std dev): 112.15(0.00)
## Number of equivalent replicates : 3106.55
##
## Deviance Information Criterion (DIC) ...............: 1844406.33
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 112.15
##
## Watanabe-Akaike information criterion (WAIC) ...: 1844409.61
## Effective number of parameters .................: 115.38
##
## Marginal log-Likelihood: -922564.07
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedSpace and Non-Parametric Time and Space-Time Interaction (Type I)
formula.T1 = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif)))
#theta1 = Model.T1$internal.summary.hyperpar$mean
theta1 = c(-2.430207, -1.077162, 4.345397, 1.679594, 1.157008)
Model.T1 = inla(formula.T1,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m3 = c(Model.T1$dic$dic, Model.T1$waic$waic)
summary(Model.T1)##
## Call:
## c("inla(formula = formula1, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 2.4554 348.0650 19.4889 370.0093
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode kld
## (Intercept) 19.5522 0.0085 19.5354 19.5522 19.5689 19.5522 0
##
## Random effects:
## Name Model
## Region Generic1 model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0880 0.0002 0.0877 0.0880
## Precision for Region 0.3434 0.0452 0.2740 0.3364
## Beta for Region 0.9864 0.0050 0.9741 0.9874
## Precision for Year1 5.5393 1.4600 3.3732 5.2947
## Precision for ID.Region.Yr 3.1921 0.2807 2.8225 3.1340
## 0.975quant mode
## Precision for the Gaussian observations 0.0883 0.0880
## Precision for Region 0.4496 0.3183
## Beta for Region 0.9933 0.9890
## Precision for Year1 9.0624 4.8229
## Precision for ID.Region.Yr 3.8744 2.9278
##
## Expected number of effective parameters(std dev): 1853.94(0.00)
## Number of equivalent replicates : 187.92
##
## Deviance Information Criterion (DIC) ...............: 1837326.90
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1853.92
##
## Watanabe-Akaike information criterion (WAIC) ...: 1837325.75
## Effective number of parameters .................: 1839.31
##
## Marginal log-Likelihood: -919790.62
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedSpatio-temporal with Regional Effect
Effect for Upper and Lower Peninsula
formula.T1r = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(Pen,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif)))
#theta1 = Model.T1r$internal.summary.hyperpar$mean
theta1 = c(-2.400136, -1.269618, 3.594011, 4.210061, 1.394595, 1.690612)
Model.T1r = inla(formula.T1r,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m4 = c(Model.T1r$dic$dic, Model.T1r$waic$waic)
summary(Model.T1r)##
## Call:
## c("inla(formula = formula1reg, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 2.7320 818.1981 20.2657 841.1958
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode kld
## (Intercept) 19.5551 0.0352 19.4861 19.5551 19.6241 19.5551 0
##
## Random effects:
## Name Model
## Region Generic1 model
## Pen IID model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0880 0.0002 0.0876 0.0880
## Precision for Region 0.3064 0.0462 0.2329 0.3000
## Beta for Region 0.9595 0.0240 0.8977 0.9651
## Precision for Pen 5.4601 1.2249 3.3074 5.3863
## Precision for Year1 7.9033 4.7639 3.1856 6.5259
## Precision for ID.Region.Yr 3.3953 0.1491 3.0841 3.4044
## 0.975quant mode
## Precision for the Gaussian observations 0.0885 0.0879
## Precision for Region 0.4133 0.2849
## Beta for Region 0.9889 0.9746
## Precision for Pen 8.0961 5.2614
## Precision for Year1 20.5417 4.7107
## Precision for ID.Region.Yr 3.6655 3.4418
##
## Expected number of effective parameters(std dev): 1790.55(0.00)
## Number of equivalent replicates : 194.57
##
## Deviance Information Criterion (DIC) ...............: 1837305.16
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1790.60
##
## Watanabe-Akaike information criterion (WAIC) ...: 1837302.20
## Effective number of parameters .................: 1775.36
##
## Marginal log-Likelihood: -919743.22
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedSpatio-temporal with Regional Effect and the NW12
Effect for Upper and Lower Peninsula and the “Northwest 12” counties subject to APRs.
formula.T1apr = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif)))
#theta1 = Model.T1apr$internal.summary.hyperpar$mean
theta1 = c(-2.400136, -1.269618, 3.594011, 4.210061, 1.394595, 1.690612)
Model.T1apr = inla(formula.T1apr,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m5 = c(Model.T1apr$dic$dic, Model.T1apr$waic$waic)
summary(Model.T1apr)##
## Call:
## c("inla(formula = formula1apr, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 2.7417 787.3496 18.0695 808.1608
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode kld
## (Intercept) 19.5752 0.0409 19.4948 19.5752 19.6555 19.5752 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0880 0.0002 0.0876 0.0880
## Precision for Region 0.3866 0.0662 0.2579 0.3878
## Beta for Region 0.9739 0.0264 0.9042 0.9825
## Precision for RegAPR 3.0388 0.5967 1.9885 3.0013
## Precision for Year1 5.0310 1.5435 2.6572 4.8129
## Precision for ID.Region.Yr 3.5836 0.1188 3.3918 3.5683
## 0.975quant mode
## Precision for the Gaussian observations 0.0884 0.0881
## Precision for Region 0.5138 0.3959
## Beta for Region 0.9992 0.9985
## Precision for RegAPR 4.3270 2.9374
## Precision for Year1 8.6546 4.4054
## Precision for ID.Region.Yr 3.8530 3.5140
##
## Expected number of effective parameters(std dev): 1792.04(0.00)
## Number of equivalent replicates : 194.41
##
## Deviance Information Criterion (DIC) ...............: 1837161.01
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1792.04
##
## Watanabe-Akaike information criterion (WAIC) ...: 1837158.97
## Effective number of parameters .................: 1777.68
##
## Marginal log-Likelihood: -919603.74
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedRegional Spatio-temporal with Covariates
Adding covariates to previous model.
formula1 = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
sbio1 + sbio2 + sW_PPTMax + sCORN_PLANTED + sWHEAT_PLANTED + sSOYBEANS_PLANTED +
sGrowSeas + sPrecEx10mm + sPrecEx20mm + sDSubZero + sPrec95pct + sMonMxDT +
sMMx1DPrec + sAMinT + sAMxT + sSumDur + sFrstDays + sPrecDep + sSnwDep
#theta1 = Model1cov$internal.summary.hyperpar$mean
theta1 = c(-2.400136, -1.269618, 3.594011, 4.210061, 1.394595, 1.690612)
Model1cov = inla(formula1,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m6 = c(Model1cov$dic$dic, Model1cov$waic$waic)
summary(Model1cov)##
## Call:
## c("inla(formula = formula1, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), control.mode = list(restart = TRUE, ", " theta = theta1), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 5.1680 1326.3301 18.7236 1350.2217
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode
## (Intercept) 19.5676 0.0392 19.4906 19.5676 19.6444 19.5676
## sbio1 0.5064 0.0952 0.3194 0.5064 0.6933 0.5064
## sbio2 -0.0827 0.0401 -0.1613 -0.0827 -0.0041 -0.0827
## sW_PPTMax 0.0205 0.0233 -0.0253 0.0205 0.0663 0.0205
## sCORN_PLANTED 0.1345 0.0466 0.0431 0.1345 0.2258 0.1345
## sWHEAT_PLANTED 0.1357 0.0375 0.0621 0.1357 0.2093 0.1357
## sSOYBEANS_PLANTED -0.0890 0.0433 -0.1741 -0.0890 -0.0040 -0.0890
## sGrowSeas 0.0036 0.0405 -0.0759 0.0036 0.0830 0.0036
## sPrecEx10mm 0.0155 0.0330 -0.0494 0.0155 0.0803 0.0155
## sPrecEx20mm -0.1360 0.0433 -0.2211 -0.1360 -0.0510 -0.1360
## sDSubZero -0.0047 0.0561 -0.1148 -0.0047 0.1052 -0.0047
## sPrec95pct 0.3097 0.0469 0.2176 0.3097 0.4018 0.3097
## sMonMxDT -0.0084 0.0501 -0.1068 -0.0084 0.0899 -0.0084
## sMMx1DPrec -0.1981 0.0387 -0.2741 -0.1981 -0.1221 -0.1981
## sAMinT 0.1572 0.0515 0.0560 0.1572 0.2583 0.1572
## sAMxT -0.1981 0.0640 -0.3238 -0.1981 -0.0725 -0.1981
## sSumDur -0.0580 0.0575 -0.1708 -0.0580 0.0547 -0.0580
## sFrstDays 0.0448 0.0412 -0.0361 0.0448 0.1257 0.0448
## sPrecDep -0.0432 0.0348 -0.1116 -0.0432 0.0251 -0.0432
## sSnwDep -0.2056 0.0384 -0.2810 -0.2056 -0.1303 -0.2056
## kld
## (Intercept) 0
## sbio1 0
## sbio2 0
## sW_PPTMax 0
## sCORN_PLANTED 0
## sWHEAT_PLANTED 0
## sSOYBEANS_PLANTED 0
## sGrowSeas 0
## sPrecEx10mm 0
## sPrecEx20mm 0
## sDSubZero 0
## sPrec95pct 0
## sMonMxDT 0
## sMMx1DPrec 0
## sAMinT 0
## sAMxT 0
## sSumDur 0
## sFrstDays 0
## sPrecDep 0
## sSnwDep 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0880 0.0002 0.0876 0.0880
## Precision for Region 0.4512 0.0647 0.3367 0.4468
## Beta for Region 0.9387 0.0329 0.8629 0.9436
## Precision for RegAPR 3.7026 0.8669 2.2577 3.6189
## Precision for Year1 5.9745 2.0088 2.5204 5.8585
## Precision for ID.Region.Yr 4.5722 0.2172 4.1551 4.5695
## 0.975quant mode
## Precision for the Gaussian observations 0.0884 0.0880
## Precision for Region 0.5903 0.4386
## Beta for Region 0.9863 0.9600
## Precision for RegAPR 5.6550 3.4619
## Precision for Year1 10.1814 5.5035
## Precision for ID.Region.Yr 5.0083 4.5674
##
## Expected number of effective parameters(std dev): 1676.13(0.00)
## Number of equivalent replicates : 207.86
##
## Deviance Information Criterion (DIC) ...............: 1837111.82
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1676.08
##
## Watanabe-Akaike information criterion (WAIC) ...: 1837109.79
## Effective number of parameters .................: 1663.67
##
## Marginal log-Likelihood: -919621.54
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedRegional Spatio-temporal (Reduced Model Above)
Removing weak linear effects.
formula1rcov = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
sbio1 + sbio2 + sCORN_PLANTED + sWHEAT_PLANTED + sSOYBEANS_PLANTED +
sPrecEx20mm + sPrec95pct + sMMx1DPrec + sAMinT + sAMxT + sSnwDep
#theta1 = Model.T1rcov$internal.summary.hyperpar$mean
theta1 = c(-2.4298949, -0.8306994, 3.7172451, 1.2343055, 1.9743757, 1.5678625)
Model.T1rcov = inla(formula1rcov,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m7 = c(Model.T1rcov$dic$dic, Model.T1rcov$waic$waic)
summary(Model.T1rcov)##
## Call:
## c("inla(formula = formula1rcov, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), control.mode = list(restart = TRUE, ", " theta = theta1), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 3.5779 602.6386 20.1282 626.3446
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode
## (Intercept) 19.5868 0.0396 19.5090 19.5868 19.6645 19.5868
## sbio1 0.4307 0.0866 0.2607 0.4307 0.6005 0.4307
## sbio2 -0.0927 0.0400 -0.1712 -0.0927 -0.0143 -0.0927
## sCORN_PLANTED 0.1345 0.0470 0.0422 0.1345 0.2268 0.1345
## sWHEAT_PLANTED 0.1411 0.0374 0.0676 0.1411 0.2145 0.1411
## sSOYBEANS_PLANTED -0.0885 0.0436 -0.1741 -0.0885 -0.0029 -0.0885
## sPrecEx20mm -0.1173 0.0359 -0.1878 -0.1173 -0.0468 -0.1173
## sPrec95pct 0.3038 0.0464 0.2127 0.3038 0.3949 0.3038
## sMMx1DPrec -0.2113 0.0367 -0.2833 -0.2113 -0.1393 -0.2113
## sAMinT 0.1465 0.0486 0.0511 0.1465 0.2418 0.1465
## sAMxT -0.1880 0.0569 -0.2996 -0.1880 -0.0765 -0.1880
## sSnwDep -0.2217 0.0263 -0.2734 -0.2217 -0.1700 -0.2217
## kld
## (Intercept) 0
## sbio1 0
## sbio2 0
## sCORN_PLANTED 0
## sWHEAT_PLANTED 0
## sSOYBEANS_PLANTED 0
## sPrecEx20mm 0
## sPrec95pct 0
## sMMx1DPrec 0
## sAMinT 0
## sAMxT 0
## sSnwDep 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.088 0.0002 0.0876 0.0880
## Precision for Region 0.441 0.0689 0.3221 0.4353
## Beta for Region 0.974 0.0127 0.9408 0.9773
## Precision for RegAPR 3.486 0.6042 2.4861 3.4210
## Precision for Year1 7.355 1.5006 4.6915 7.2768
## Precision for ID.Region.Yr 4.818 0.4694 4.1943 4.7216
## 0.975quant mode
## Precision for the Gaussian observations 0.0885 0.0880
## Precision for Region 0.5920 0.4238
## Beta for Region 0.9883 0.9824
## Precision for RegAPR 4.8476 3.2839
## Precision for Year1 10.5344 7.1520
## Precision for ID.Region.Yr 5.9591 4.3941
##
## Expected number of effective parameters(std dev): 1693.24(0.00)
## Number of equivalent replicates : 205.76
##
## Deviance Information Criterion (DIC) ...............: 1837108.65
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1693.18
##
## Watanabe-Akaike information criterion (WAIC) ...: 1837106.55
## Effective number of parameters .................: 1680.43
##
## Marginal log-Likelihood: -919574.45
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedRe-check for Colinearity
These checks are to avoid colinearity; cross-correlation between the El Nino index and response variable is assessed later in the script.
Colin.df = MyBeam3.tst %>%
select(sbio1, sbio2, sCORN_PLANTED, sWHEAT_PLANTED, sSOYBEANS_PLANTED,
sPrecEx20mm, sPrec95pct, sMMx1DPrec, sAMinT, sAMxT, sSnwDep, ONI.indx)
Colin.df = Colin.df[complete.cases(Colin.df),]
cor.mtest = function(mat, ...) {
mat = as.matrix(mat)
n = ncol(mat)
p.mat= matrix(NA, n, n)
diag(p.mat) = 0
for (i in 1:(n - 1)) {
for (j in (i + 1):n) {
tmp = cor.test(mat[, i], mat[, j], ...)
p.mat[i, j] = p.mat[j, i] = tmp$p.value
}}
colnames(p.mat) = rownames(p.mat) = colnames(mat)
p.mat
}
p.mat = cor.mtest(Colin.df)
col = colorRampPalette(rev(c("#BB4444", "#EE9988", "#FFFFFF", "#77AADD", "#4477AA")))
corrplot(cor(Colin.df), method="color", col=col(200),
type="upper", order="hclust",
addCoef.col = "black",
tl.col="black", tl.srt=45,
p.mat = p.mat, sig.level = 0.01, insig = "blank",
diag=FALSE)
CorCov = cor(Colin.df)
CI = colldiag(CorCov)
CI## Condition
## Index Variance Decomposition Proportions
## intercept sbio1 sbio2 sCORN_PLANTED sWHEAT_PLANTED
## 1 1.000 0.006 0.002 0.008 0.001 0.001
## 2 1.454 0.008 0.000 0.002 0.000 0.000
## 3 1.684 0.001 0.001 0.004 0.001 0.002
## 4 2.161 0.018 0.001 0.000 0.000 0.000
## 5 2.678 0.032 0.001 0.002 0.000 0.002
## 6 3.204 0.314 0.008 0.136 0.000 0.003
## 7 4.377 0.013 0.019 0.092 0.000 0.009
## 8 5.560 0.126 0.249 0.073 0.000 0.029
## 9 8.168 0.012 0.019 0.080 0.003 0.012
## 10 14.341 0.037 0.057 0.004 0.088 0.104
## 11 15.453 0.379 0.021 0.567 0.001 0.092
## 12 19.589 0.054 0.622 0.034 0.906 0.746
## sSOYBEANS_PLANTED sPrecEx20mm sPrec95pct sMMx1DPrec sAMinT sAMxT
## 1 0.001 0.002 0.000 0.000 0.001 0.001
## 2 0.000 0.009 0.006 0.013 0.000 0.001
## 3 0.001 0.001 0.000 0.000 0.010 0.015
## 4 0.000 0.000 0.001 0.000 0.000 0.006
## 5 0.000 0.005 0.002 0.028 0.001 0.001
## 6 0.001 0.044 0.002 0.002 0.000 0.001
## 7 0.000 0.213 0.000 0.084 0.000 0.011
## 8 0.000 0.003 0.017 0.002 0.006 0.094
## 9 0.012 0.007 0.116 0.194 0.308 0.201
## 10 0.636 0.309 0.278 0.311 0.036 0.152
## 11 0.331 0.356 0.506 0.279 0.626 0.434
## 12 0.017 0.051 0.071 0.086 0.010 0.084
## sSnwDep ONI.indx
## 1 0.002 0.001
## 2 0.007 0.000
## 3 0.004 0.028
## 4 0.138 0.202
## 5 0.221 0.159
## 6 0.006 0.017
## 7 0.133 0.052
## 8 0.025 0.004
## 9 0.092 0.174
## 10 0.036 0.001
## 11 0.069 0.350
## 12 0.267 0.012Adding Oceanic Nino Index
The ONI is a measure of the El Niño-Southern Oscillation.
pcprior = list(prec = list(prior="pc.prec", param = c(1, 0.0001)))
formulaONI = AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(ONI.indx,
model="rw1",
constr=TRUE,
hyper=pcprior) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
sbio1 + sbio2 + sCORN_PLANTED + sWHEAT_PLANTED + sSOYBEANS_PLANTED +
sPrecEx20mm + sPrec95pct + sMMx1DPrec + sAMinT + sAMxT + sSnwDep
#theta1 = Model.T1oni$internal.summary.hyperpar$mean
theta1 = c(-2.4548244, -0.9312593, 2.7321389, 1.4852687, 0.6983412, 1.5902152, 1.5257093)
Model.T1oni = inla(formulaONI,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = MyBeam3.tst)dic.m8 = c(Model.T1oni$dic$dic, Model.T1oni$waic$waic)
summary(Model.T1oni)##
## Call:
## c("inla(formula = formulaONI, family = \"gaussian\", data = MyBeam3.tst, ", " verbose = TRUE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"))")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 3.9137 2543.2267 22.1809 2569.3212
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode
## (Intercept) 19.5311 0.0383 19.4558 19.5311 19.6063 19.5311
## sbio1 0.4928 0.0919 0.3123 0.4928 0.6732 0.4928
## sbio2 -0.0876 0.0401 -0.1663 -0.0876 -0.0089 -0.0876
## sCORN_PLANTED 0.1337 0.0465 0.0424 0.1337 0.2250 0.1337
## sWHEAT_PLANTED 0.1473 0.0369 0.0749 0.1473 0.2196 0.1473
## sSOYBEANS_PLANTED -0.0905 0.0432 -0.1753 -0.0905 -0.0059 -0.0905
## sPrecEx20mm -0.1198 0.0370 -0.1925 -0.1198 -0.0471 -0.1198
## sPrec95pct 0.3239 0.0473 0.2310 0.3239 0.4167 0.3239
## sMMx1DPrec -0.2165 0.0374 -0.2899 -0.2165 -0.1432 -0.2165
## sAMinT 0.1540 0.0526 0.0508 0.1540 0.2571 0.1540
## sAMxT -0.2276 0.0612 -0.3477 -0.2276 -0.1077 -0.2276
## sSnwDep -0.2268 0.0259 -0.2777 -0.2268 -0.1759 -0.2268
## kld
## (Intercept) 0
## sbio1 0
## sbio2 0
## sCORN_PLANTED 0
## sWHEAT_PLANTED 0
## sSOYBEANS_PLANTED 0
## sPrecEx20mm 0
## sPrec95pct 0
## sMMx1DPrec 0
## sAMinT 0
## sAMxT 0
## sSnwDep 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## ONI.indx RW1 model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0859 0.0001 0.0857 0.0859
## Precision for Region 0.3966 0.0456 0.3203 0.3917
## Beta for Region 0.9352 0.0248 0.8719 0.9409
## Precision for RegAPR 4.4683 0.6689 3.1325 4.4945
## Precision for ONI.indx 2.0671 0.4673 1.1616 2.0796
## Precision for Year1 5.0120 1.0167 3.0819 5.0132
## Precision for ID.Region.Yr 4.6012 0.1607 4.2570 4.6146
## 0.975quant mode
## Precision for the Gaussian observations 0.0862 0.0858
## Precision for Region 0.4992 0.3794
## Beta for Region 0.9669 0.9511
## Precision for RegAPR 5.7141 4.6289
## Precision for ONI.indx 2.9346 2.1372
## Precision for Year1 7.0019 5.0740
## Precision for ID.Region.Yr 4.8838 4.6674
##
## Expected number of effective parameters(std dev): 1678.67(0.00)
## Number of equivalent replicates : 207.54
##
## Deviance Information Criterion (DIC) ...............: 1836912.50
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1678.77
##
## Watanabe-Akaike information criterion (WAIC) ...: 1836867.31
## Effective number of parameters .................: 1623.53
##
## Marginal log-Likelihood: -919534.77
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedLag Data
Modifying best model above (Model.T1oni) for a one and two-year lag.
#Copy data
MyBeam4.tst = MyBeam3.tst
#Need to lag by county-year (one county and year at a time)
CntyYr.lvls = levels(factor(MyBeam4.tst$Region.Yr))
for(i in 1:length(CntyYr.lvls)){
CntyYr.tmp = MyBeam4.tst %>% filter(Region.Yr == CntyYr.lvls[i])
CntyYr.tmp$Lag1.AvgBeam = lag(CntyYr.tmp$AvgBeam, 1) #1 year lag
CntyYr.tmp$Lag2.AvgBeam = lag(CntyYr.tmp$AvgBeam, 2)#2 year lag
if(i == 1){L2.MyBeam4.tst = CntyYr.tmp
} else{L2.MyBeam4.tst = rbind(L2.MyBeam4.tst, CntyYr.tmp)}
}
#save(list=c("L2.MyBeam4.tst", "Q.Leroux", "g", "sdunif", "lunif"), file="C:/Users/humph173/Documents/Michigan_State/SLP_Beam_Diam/HPCC/Aug2519/Lag2yr.RData")
#save(list=c("L2.MyBeam4.tst", "Q.Leroux", "g", "sdunif", "lunif"), file="C:/Users/humph173/Documents/Michigan_State/SLP_Beam_Diam/HPCC/Aug2519/Lag1_2yr.RData")One Year Lag
pcprior = list(prec = list(prior="pc.prec", param = c(1, 0.0001)))
formulaONI = Lag1.AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(ONI.indx,
model="rw1",
constr=TRUE,
hyper=pcprior) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
sbio1 + sbio2 + sCORN_PLANTED + sWHEAT_PLANTED + sSOYBEANS_PLANTED +
sPrecEx20mm + sPrec95pct + sMMx1DPrec + sAMinT + sAMxT + sSnwDep
#theta1 = Model.T1oni$internal.summary.hyperpar$mean
theta1 = c(-2.4548244, -0.9312593, 2.7321389, 1.4852687, 0.6983412, 1.5902152, 1.5257093)
Model.T1oni.l1 = inla(formulaONI,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = L2.MyBeam4.tst)dic.m9 = c(Model.T1oni.l1$dic$dic, Model.T1oni.l1$waic$waic)
summary(Model.T1oni.l1)##
## Call:
## c("inla(formula = formulaONI, family = \"gaussian\", data = L2.MyBeam4.tst, ", " verbose = FALSE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 4.9514 2692.3701 27.2380 2724.5596
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode
## (Intercept) 19.5679 0.0469 19.4759 19.5679 19.6598 19.5679
## sbio1 0.4539 0.0946 0.2681 0.4539 0.6396 0.4539
## sbio2 -0.1017 0.0403 -0.1808 -0.1017 -0.0226 -0.1017
## sCORN_PLANTED 0.1421 0.0468 0.0502 0.1421 0.2340 0.1421
## sWHEAT_PLANTED 0.1463 0.0375 0.0727 0.1463 0.2199 0.1463
## sSOYBEANS_PLANTED -0.0853 0.0434 -0.1705 -0.0853 -0.0002 -0.0853
## sPrecEx20mm -0.1194 0.0385 -0.1949 -0.1194 -0.0439 -0.1194
## sPrec95pct 0.3265 0.0490 0.2302 0.3265 0.4227 0.3265
## sMMx1DPrec -0.2196 0.0388 -0.2958 -0.2196 -0.1435 -0.2196
## sAMinT 0.1579 0.0552 0.0496 0.1579 0.2662 0.1579
## sAMxT -0.2185 0.0643 -0.3447 -0.2185 -0.0924 -0.2185
## sSnwDep -0.2265 0.0267 -0.2789 -0.2265 -0.1741 -0.2265
## kld
## (Intercept) 0
## sbio1 0
## sbio2 0
## sCORN_PLANTED 0
## sWHEAT_PLANTED 0
## sSOYBEANS_PLANTED 0
## sPrecEx20mm 0
## sPrec95pct 0
## sMMx1DPrec 0
## sAMinT 0
## sAMxT 0
## sSnwDep 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## ONI.indx RW1 model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0881 0.0002 0.0878 0.0881
## Precision for Region 0.5355 0.0628 0.4131 0.5360
## Beta for Region 0.9933 0.0084 0.9714 0.9959
## Precision for RegAPR 2.5870 0.5438 1.6783 2.5333
## Precision for ONI.indx 2.7913 0.7397 1.5823 2.7119
## Precision for Year1 4.9197 1.3616 2.8266 4.7212
## Precision for ID.Region.Yr 4.4262 0.1747 4.1194 4.4118
## 0.975quant mode
## Precision for the Gaussian observations 0.0885 0.0881
## Precision for Region 0.6583 0.5417
## Beta for Region 0.9993 0.9982
## Precision for RegAPR 3.8090 2.4305
## Precision for ONI.indx 4.4758 2.5608
## Precision for Year1 8.1324 4.3486
## Precision for ID.Region.Yr 4.8033 4.3667
##
## Expected number of effective parameters(std dev): 1714.83(0.00)
## Number of equivalent replicates : 201.67
##
## Deviance Information Criterion (DIC) ...............: 1823074.52
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1714.88
##
## Watanabe-Akaike information criterion (WAIC) ...: 1823072.72
## Effective number of parameters .................: 1702.08
##
## Marginal log-Likelihood: -912633.02
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedTwo Year Lag
pcprior = list(prec = list(prior="pc.prec", param = c(1, 0.0001)))
formulaONI = Lag2.AvgBeam ~ f(Region,
model="generic1",
Cmatrix=Q.Leroux,
constr=TRUE,
hyper=list(prec=list(prior=sdunif),
beta=list(prior=lunif))) +
f(RegAPR,
model="iid",
hyper=reg.prior,
replicate=Year2,
constr = TRUE) +
f(ONI.indx,
model="rw1",
constr=TRUE,
hyper=pcprior) +
f(Year1,
model="rw1",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
f(ID.Region.Yr,
model="iid",
constr=TRUE,
hyper=list(prec=list(prior=sdunif))) +
sbio1 + sbio2 + sCORN_PLANTED + sWHEAT_PLANTED + sSOYBEANS_PLANTED +
sPrecEx20mm + sPrec95pct + sMMx1DPrec + sAMinT + sAMxT + sSnwDep
#theta1 = Model.T1oni$internal.summary.hyperpar$mean
theta1 = c(-2.4548244, -0.9312593, 2.7321389, 1.4852687, 0.6983412, 1.5902152, 1.5257093)
Model.T1oni.l2 = inla(formulaONI,
family = "gaussian",
verbose = FALSE,
control.predictor = list(
compute = TRUE,
link = 1),
control.mode = list(restart = TRUE, theta = theta1),
control.inla = list(strategy = "gaussian",
int.strategy = "eb"),
control.compute = list(cpo = TRUE, waic=TRUE, dic=TRUE),
data = L2.MyBeam4.tst)dic.m10 = c(Model.T1oni.l2$dic$dic, Model.T1oni.l2$waic$waic)
summary(Model.T1oni.l2)##
## Call:
## c("inla(formula = formulaONI, family = \"gaussian\", data = L2.MyBeam4.tst, ", " verbose = FALSE, control.compute = list(cpo = TRUE, waic = TRUE, ", " dic = TRUE), control.predictor = list(compute = TRUE, ", " link = 1), control.inla = list(strategy = \"gaussian\", ", " int.strategy = \"eb\"), num.threads = 6)")
##
## Time used:
## Pre-processing Running inla Post-processing Total
## 5.3643 2164.9179 27.7029 2197.9851
##
## Fixed effects:
## mean sd 0.025quant 0.5quant 0.975quant mode
## (Intercept) 19.5657 0.0432 19.4810 19.5657 19.6504 19.5657
## sbio1 0.4584 0.0971 0.2677 0.4584 0.6488 0.4584
## sbio2 -0.1049 0.0417 -0.1869 -0.1049 -0.0230 -0.1049
## sCORN_PLANTED 0.1399 0.0485 0.0446 0.1399 0.2351 0.1399
## sWHEAT_PLANTED 0.1458 0.0389 0.0695 0.1458 0.2221 0.1458
## sSOYBEANS_PLANTED -0.0844 0.0451 -0.1729 -0.0844 0.0039 -0.0844
## sPrecEx20mm -0.1205 0.0396 -0.1982 -0.1205 -0.0428 -0.1205
## sPrec95pct 0.3313 0.0505 0.2321 0.3313 0.4304 0.3313
## sMMx1DPrec -0.2201 0.0400 -0.2986 -0.2201 -0.1417 -0.2201
## sAMinT 0.1528 0.0566 0.0416 0.1528 0.2639 0.1528
## sAMxT -0.2203 0.0659 -0.3497 -0.2203 -0.0910 -0.2203
## sSnwDep -0.2318 0.0275 -0.2859 -0.2318 -0.1778 -0.2318
## kld
## (Intercept) 0
## sbio1 0
## sbio2 0
## sCORN_PLANTED 0
## sWHEAT_PLANTED 0
## sSOYBEANS_PLANTED 0
## sPrecEx20mm 0
## sPrec95pct 0
## sMMx1DPrec 0
## sAMinT 0
## sAMxT 0
## sSnwDep 0
##
## Random effects:
## Name Model
## Region Generic1 model
## RegAPR IID model
## ONI.indx RW1 model
## Year1 RW1 model
## ID.Region.Yr IID model
##
## Model hyperparameters:
## mean sd 0.025quant 0.5quant
## Precision for the Gaussian observations 0.0882 0.0002 0.0878 0.0882
## Precision for Region 0.4687 0.0782 0.3159 0.4704
## Beta for Region 0.9302 0.0659 0.7463 0.9510
## Precision for RegAPR 3.9946 1.2276 2.4176 3.7149
## Precision for ONI.indx 3.5467 1.5271 1.1938 3.3626
## Precision for Year1 5.8398 2.7476 2.6194 5.1457
## Precision for ID.Region.Yr 4.1113 0.1789 3.8374 4.0840
## 0.975quant mode
## Precision for the Gaussian observations 0.0886 0.0882
## Precision for Region 0.6180 0.4811
## Beta for Region 0.9912 0.9772
## Precision for RegAPR 7.0982 3.1752
## Precision for ONI.indx 7.0242 2.8457
## Precision for Year1 12.9807 4.0958
## Precision for ID.Region.Yr 4.5252 3.9859
##
## Expected number of effective parameters(std dev): 1745.59(0.00)
## Number of equivalent replicates : 196.64
##
## Deviance Information Criterion (DIC) ...............: 1809405.09
## Deviance Information Criterion (DIC, saturated) ....: NULL
## Effective number of parameters .....................: 1745.53
##
## Watanabe-Akaike information criterion (WAIC) ...: 1809404.63
## Effective number of parameters .................: 1733.40
##
## Marginal log-Likelihood: -905783.93
## CPO and PIT are computed
##
## Posterior marginals for linear predictor and fitted values computedCompare Models
Models = c("Model1", "Model2",
"Model3", "Model4",
"Model5", "Model6",
"Model7", "Model8",
"Model9", "Model10")
Type = c("Space Only",
"Space + Time",
"Space + Time + Interaction",
"Space + Time + Interaction + Region",
"Model above + NW12 (APR)",
"Model above + Linear Covariates",
"Model Above Reduced",
"Model Above + El Nino",
"Model Above + 1yr Lag",
"Two Models Above + 2yr Lag")
#Deviance information criteria
DICs = c(dic.m1[1], dic.m2[1], dic.m3[1],
dic.m4[1], dic.m5[1], dic.m6[1],
dic.m7[1], dic.m8[1], dic.m9[1],
dic.m10[1])
#Watanabe-Akaike information criteria
WAICs = c(dic.m1[2], dic.m2[2], dic.m3[2],
dic.m4[2], dic.m5[2], dic.m6[2],
dic.m7[2], dic.m8[2], dic.m9[2],
dic.m10[2])
Model_mets = as.data.frame(cbind(Model = Models,
DIC = round(DICs, 0),
WAIC = round(WAICs, 0),
Type = Type))kable(Model_mets) %>%
kable_styling("striped", full_width = F) %>%
row_spec(0, font_size = 20) | Model | DIC | WAIC | Type |
|---|---|---|---|
| Model1 | 2793712 | 2793999 | Space Only |
| Model2 | 1844406 | 1844410 | Space + Time |
| Model3 | 1837327 | 1837326 | Space + Time + Interaction |
| Model4 | 1837305 | 1837302 | Space + Time + Interaction + Region |
| Model5 | 1837161 | 1837159 | Model above + NW12 (APR) |
| Model6 | 1837112 | 1837110 | Model above + Linear Covariates |
| Model7 | 1837109 | 1837107 | Model Above Reduced |
| Model8 | 1836912 | 1836867 | Model Above + El Nino |
| Model9 | 1823075 | 1823073 | Model Above + 1yr Lag |
| Model10 | 1809405 | 1809405 | Two Models Above + 2yr Lag |
Statewide Temporal Effect
mic.d = as.data.frame(Model.T1oni.l2$summary.random$Year1)[,c(1:6)]
names(mic.d) = c("ID", "Mean", "sd", "Q025", "Q50", "Q975")
mic.d$YLab = paste(levels(factor(MyBeam3$year)))
YLab = paste(mic.d$YLab)
ggplot(mic.d, aes(ID, Mean), lab) +
geom_smooth(col = "black",
linetype= "solid",
method = "loess",
span = 0.4,
se = FALSE,
lwd = 1) +
geom_smooth(data = mic.d, aes(ID, Q025),
col = "grey",
method = "loess",
span = 0.4,
se = FALSE,
linetype= "dashed") +
geom_smooth(data = mic.d, aes(ID, Q975),
col = "grey",
method = "loess",
span = 0.4,
se = FALSE,
linetype= "dashed") +
geom_hline(yintercept = 0,
linetype = "solid",
col = "darkgray",
size = 0.5) +
xlab(" ") +
ylab("Statewide Temporal Effect") +
theme_classic() +
scale_y_continuous(breaks = seq(-0.5, 1.75, 0.5)) +
scale_x_continuous(breaks = 0:30,labels=paste0(mic.d$YLab)) +
theme(axis.text=element_text(size=16),
strip.text = element_text(face="bold", size = 20),
axis.title.y = element_text(face="bold", size = 20),
axis.text.x = element_text(face="bold", size=14, vjust=0.5,
hjust=1, angle=90),
axis.title.x = element_text(face="bold", size = 20))
Oceanic Nino Index
mic.df = as.data.frame(Model.T1oni.l2$summary.random$ONI.indx)[,c(1:6)]
names(mic.df) = c("ID", "Mean", "sd", "Q025", "Q50", "Q975")
ONI.plt = MyBeam4.tst[,c("Year", "Month", "ONI.indx")]
ONI.plt$Mean = with(mic.df,
Mean[match(
ONI.plt$ONI.indx,
ID)])
ONI.plt$Q025 = with(mic.df,
Q025[match(
ONI.plt$ONI.indx,
ID)])
ONI.plt$Q975 = with(mic.df,
Q975[match(
ONI.plt$ONI.indx,
ID)])
plt.df = as.data.frame(
ONI.plt %>%
group_by(Year, Month) %>%
summarise(Mean = mean(Mean),
MeanL = mean(Q025),
MeanH = mean(Q975)))
YLab = paste(levels(factor(MyBeam3$year)))
ggplot(plt.df, aes(Year, Mean, color = Mean)) +
geom_smooth(aes(col = ..y..),
linetype= "solid",
method = "loess",
span = 0.3,
se = FALSE,
lwd = 2) +
scale_colour_gradient2(low = "blue",
high = "red",
midpoint=0) +
geom_smooth(data = plt.df, aes(Year, MeanL),
col = "grey",
method = "loess",
span = 0.3,
se = FALSE,
linetype= "dashed") +
geom_smooth(data = plt.df, aes(Year, MeanH),
col = "grey",
method = "loess",
span = 0.3,
se = FALSE,
linetype= "dashed") +
geom_hline(yintercept = 0,
linetype = "solid",
col = "darkgray",
size = 0.5) +
xlab(" ") +
ylab("Oceanic Niño Index Effect") +
theme_classic() +
scale_x_continuous(breaks = seq(1987, 2017, 1), labels=paste0(YLab)) +
theme(axis.text=element_text(size=16),
strip.text = element_text(face="bold", size = 20),
axis.title.y = element_text(face="bold", size = 20),
legend.position="none",
axis.text.x = element_text(face="bold", size=14, vjust=0.5,
hjust=1, angle=90),
axis.title.x = element_text(face="bold", size = 20))
Compare Temporal Effect with ONI
ggplot(plt.df, aes(Year, Mean, color = Mean)) +
geom_smooth(aes(col = ..y..),
linetype= "solid",
method = "loess",
span = 0.3,
se = FALSE,
lwd = 2) +
scale_colour_gradient2(low = "blue",
high = "red",
midpoint=0) +
geom_smooth(data = mic.d, aes(as.numeric(YLab), Mean),
col = "black",
method = "loess",
span = 0.3,
se = FALSE,
linetype= "solid") +
geom_hline(yintercept = 0,
linetype = "solid",
col = "darkgray",
size = 0.5) +
xlab(" ") +
ylab("Effect Size") +
theme_classic() +
scale_y_continuous(breaks = seq(-0.5, 1.75, 0.5)) +
scale_x_continuous(breaks = seq(1987, 2017, 1), labels=paste0(YLab)) +
theme(axis.text=element_text(size=16),
strip.text = element_text(face="bold", size = 20),
axis.title.y = element_text(face="bold", size = 20),
legend.position="none",
axis.text.x = element_text(face="bold", size=14, vjust=0.5,
hjust=1, angle=90),
axis.title.x = element_text(face="bold", size = 20))
Region Effect
Region.plt = as.data.frame(Model.T1oni.l2$summary.random$RegAPR)
names(Region.plt) = c("ID", "Mean", "sd", "Q025", "Q50", "Q975", "mode", "kld")
Region.plt$ID2 = factor(rep(c("Northwest 12", "Lower Peninsula", "Upper Peninsula"),31))
Region.plt$ID2 = factor(Region.plt$ID2, levels = c("Upper Peninsula", "Lower Peninsula", "Northwest 12"))
Region.plt$Year = as.integer(rep(Yr.lv, each=3))
ggplot(Region.plt, aes(x=Year, y=Mean)) +
geom_point(size=2, pch=19, col = "red") +
geom_linerange(aes(ymin=Q025, ymax=Q975), colour="black") +
geom_hline(yintercept = 0,
linetype = "dotted",
colour = "darkgray",
size = 0.5) +
scale_x_continuous(breaks = seq(1987, 2017, 1), labels=paste0(YLab)) +
facet_wrap(~ID2, ncol=1) +
theme_classic() +
xlab(NULL) +
ylab("Regional Grouping Effect ") +
theme(axis.title.y = element_text(face="bold", size=18),
strip.text.x = element_text(face="bold", size = 14, colour = "black"),
axis.text.y = element_text(face="bold", size=12),
axis.text.x = element_text(face="bold", size=14, vjust=0.5,
hjust=1, angle=90))
Linear Effects
Fixed.df = Model.T1oni.l2$summary.fixed[,c(1:3,5)]
names(Fixed.df) = c("Mean", "SD", "Q025", "Q975")
Fixed.df = round(Fixed.df[,1:4], 3)kable(Fixed.df) %>%
kable_styling("striped", full_width = F) %>%
row_spec(0, font_size = 20) %>%
column_spec(1, bold = T)| Mean | SD | Q025 | Q975 | |
|---|---|---|---|---|
| (Intercept) | 19.566 | 0.043 | 19.481 | 19.650 |
| sbio1 | 0.458 | 0.097 | 0.268 | 0.649 |
| sbio2 | -0.105 | 0.042 | -0.187 | -0.023 |
| sCORN_PLANTED | 0.140 | 0.049 | 0.045 | 0.235 |
| sWHEAT_PLANTED | 0.146 | 0.039 | 0.070 | 0.222 |
| sSOYBEANS_PLANTED | -0.084 | 0.045 | -0.173 | 0.004 |
| sPrecEx20mm | -0.120 | 0.040 | -0.198 | -0.043 |
| sPrec95pct | 0.331 | 0.051 | 0.232 | 0.430 |
| sMMx1DPrec | -0.220 | 0.040 | -0.299 | -0.142 |
| sAMinT | 0.153 | 0.057 | 0.042 | 0.264 |
| sAMxT | -0.220 | 0.066 | -0.350 | -0.091 |
| sSnwDep | -0.232 | 0.028 | -0.286 | -0.178 |
Spatial Effect Random Field Density
LU.tab = as.data.frame(Model.T1oni.l2$summary.random$Region[, c(1:2,4,6)])
names(LU.tab) = c("ID", "Mean", "Q025", "Q975")
LU.tab$County = LookUp.tab$NAME
Counties2 = Counties
Domain.r = raster(res=0.01,
ext = extent(Counties))
Counties2$RE = LU.tab$Mean
Region.mean.r = rasterize(Counties2,
Domain.r,
field = "RE",
background = NA)
Counties2$Q025 = LU.tab$Q025
Region.Q025.r = rasterize(Counties2,
Domain.r,
field = "Q025",
background = NA)
Counties2$Q975 = LU.tab$Q975
Region.Q975.r = rasterize(Counties2,
Domain.r,
field = "Q975",
background = NA)
MRF.stk = stack(Region.Q025.r, Region.mean.r, Region.Q975.r)
names(MRF.stk) = c("A", "B", "C")Plot Spatial Random Field
mCols = c("#A50026", "#D73027", "#F46D43", "#FDAE61", "#E0F3F8", "#ABD9E9", "#74ADD1", "#4575B4","#313695")
rng = seq(-3.3, 3.3, 0.01)
cr = colorRampPalette(c(rev(mCols)),
bias = 1, space = "rgb")
levelplot(MRF.stk,
margin = FALSE,
maxpixels = 1e5,
layout=c(3, 1),
xlab = " ", ylab = NULL,
names.attr= c("A","B","C"),
col.regions = cr, at = rng,
colorkey = list(space = "bottom"),
par.settings = list(axis.line = list(col = 'transparent'),
strip.background = list(col = 'transparent'),
strip.border = list(col = 'transparent')),
par.strip.text = list(cex=1, fontface='bold', col="black"),
scales = list(draw=FALSE)) +
latticeExtra::layer(sp.polygons(Counties, col = "darkgray", lwd = 0.5)) 
Compare Model.sp and Model.T1oni.l2 (Spatial Random Fields)
names(MRF.stk) = c("D", "E", "F")
MRF.stk2 = stack(MRF.stk2, MRF.stk)
rng = seq(-35, 5, 0.01)
cr = colorRampPalette(c(rev(mCols)),
bias = 0.20, space = "rgb")
levelplot(MRF.stk2,
margin = FALSE,
maxpixels = 1e5,
layout=c(3, 2),
xlab = " ", ylab = NULL,
names.attr= c("A","B","C", "D", "E", "F"),
col.regions = cr, at = rng,
colorkey = list(labels=list(at=seq(-35, 5, 5),
labels=c("-35","-30","-25","-20","-15","-10","-5","0","5"),
cex=1.5),
labels=list(cex=12),
space = "bottom"),
par.settings = list(axis.line = list(col = 'transparent'),
strip.background = list(col = 'transparent'),
strip.border = list(col = 'transparent')),
par.strip.text = list(cex=1, fontface='bold', col="black"),
scales = list(draw=FALSE)) +
latticeExtra::layer(sp.polygons(Counties, col = "darkgray", lwd = 0.5)) 
Space-Time Interaction Mean Random Field Density
LU.tab = as.data.frame(Model.T1oni.l2$summary.random$ID.Region.Yr[, c(1:2,4,6)])
names(LU.tab) = c("ID", "Mean", "Q025", "Q975")
LU.tab$County = with(MyBeam4.tst,
County[match(
LU.tab$ID,
ID.Region.Yr)])
LU.tab$Year = with(MyBeam4.tst,
Year[match(
LU.tab$ID,
ID.Region.Yr)])
Counties2 = Counties
for(i in 1:length(Yr.lv)){
tmp.lu = LU.tab %>%
filter(Year == Yr.lv[i])
Counties2$RE = with(tmp.lu,
Mean[match(
Counties2$NAME,
County)])
Tmp.r = rasterize(Counties2,
Domain.r,
field = "RE",
background = NA)
if(i == 1){Int.stk = Tmp.r}
else{Int.stk = stack(Int.stk, Tmp.r)}
}
names(Int.stk) = paste("Year.", Yr.lv, sep="")Plot Interaction Fields
Rng.stk = range(values(Int.stk), na.rm=T)
rng = seq(Rng.stk[1]-0.02, Rng.stk[2]+0.02, 0.01)
mCols = c("#A50026", "#D73027", "#F46D43", "#FDAE61", "#E0F3F8", "#ABD9E9", "#74ADD1", "#4575B4","#313695")
#brewer.pal(11, "RdBu")
cr = colorRampPalette(c(rev(mCols)),
bias = 1, space = "rgb")
levelplot(Int.stk,
margin = FALSE,
maxpixels = 1e5,
layout=c(5, 7),
xlab = " ", ylab = NULL,
names.attr= Yr.lv,
col.regions = cr, at = rng,
colorkey = FALSE,
par.settings = list(axis.line = list(col = 'transparent'),
strip.background = list(col = 'transparent'),
strip.border = list(col = 'transparent')),
par.strip.text = list(cex=1, fontface='bold', col="black"),
scales = list(draw=FALSE)) +
latticeExtra::layer(sp.polygons(Counties, col = "darkgray", lwd = 0.5)) 
Fit Values
MyBeam4.tst$Fit = Model.T1oni.l2$summary.fitted.values$mean
Counties2 = Counties
for(i in 1:length(Yr.lv)){
tmp.lu = MyBeam4.tst %>%
filter(Year == Yr.lv[i])
Counties2$Fit = with(tmp.lu,
Fit[match(
Counties2$NAME,
County)])
Tmp.r = rasterize(Counties2,
Domain.r,
field = "Fit",
background = NA)
if(i == 1){Fit.stk = Tmp.r}
else{Fit.stk = stack(Fit.stk, Tmp.r)}
}
names(Fit.stk) = paste("Yr.", Yr.lv, sep="")Plot Fit Values
Rng.stk = range(values(Fit.stk), na.rm=T)
rng = seq(14, 24, 0.01)
mCols = brewer.pal(9, "YlOrBr")
#brewer.pal(11, "YlOrBr")
cr = colorRampPalette(c(mCols),
bias = 1, space = "rgb")
levelplot(Fit.stk,
margin = FALSE,
maxpixels = 1e5,
layout=c(5, 7),
xlab = " ", ylab = NULL,
names.attr= Yr.lv,
col.regions = cr, at = rng,
colorkey = list(labels=list(at=c(14, 16, 18, 20, 22, 24),
labels=c("14 mm", "16 mm", "18 mm", "20 mm", "22 mm", "24 mm"), cex=1.5),
labels=list(cex=12),
space = "right"),
par.settings = list(axis.line = list(col = 'transparent'),
strip.background = list(col = 'transparent'),
strip.border = list(col = 'transparent')),
par.strip.text = list(cex=1, fontface='bold', col="black"),
scales = list(draw=FALSE)) +
latticeExtra::layer(sp.polygons(Counties, col = "darkgray", lwd = 0.5)) 
Session Info
sessionInfo()## R version 3.5.3 (2019-03-11)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 10 x64 (build 17134)
##
## Matrix products: default
##
## locale:
## [1] LC_COLLATE=English_United States.1252
## [2] LC_CTYPE=English_United States.1252
## [3] LC_MONETARY=English_United States.1252
## [4] LC_NUMERIC=C
## [5] LC_TIME=English_United States.1252
##
## attached base packages:
## [1] grid stats graphics grDevices utils datasets methods
## [8] base
##
## other attached packages:
## [1] perturb_2.10 astsa_1.8 lubridate_1.7.4
## [4] lettercase_0.13.1 kableExtra_1.1.0 stringr_1.4.0
## [7] readr_1.3.1 data.table_1.12.0 tseries_0.10-46
## [10] dismo_1.1-4 dplyr_0.8.0.1 Rmisc_1.5
## [13] plyr_1.8.4 xtable_1.8-3 gridExtra_2.3
## [16] GISTools_0.7-4 ggthemes_4.1.0 spdep_1.0-2
## [19] sf_0.7-3 spData_0.3.0 mapproj_1.2.6
## [22] maptools_0.9-5 rgeos_0.4-2 rasterVis_0.45
## [25] latticeExtra_0.6-28 RColorBrewer_1.1-2 lattice_0.20-38
## [28] corrplot_0.84 factoextra_1.0.5 ggplot2_3.1.0
## [31] seas_0.5-2 MASS_7.3-51.1 climdex.pcic_1.1-9.1
## [34] PCICt_0.5-4.1 rnoaa_0.8.4 missMDA_1.14
## [37] FactoMineR_1.41 fields_9.6 maps_3.3.0
## [40] spam_2.2-2 dotCall64_1.0-0 raster_2.8-19
## [43] reshape2_1.4.3 rgdal_1.4-3 INLA_18.07.12
## [46] sp_1.3-1 Matrix_1.2-17 rgl_0.100.19
## [49] knitr_1.22
##
## loaded via a namespace (and not attached):
## [1] backports_1.1.3 lazyeval_0.2.2
## [3] splines_3.5.3 crosstalk_1.0.0
## [5] digest_0.6.18 foreach_1.4.4
## [7] htmltools_0.3.6 gdata_2.18.0
## [9] magrittr_1.5 cluster_2.0.7-1
## [11] doParallel_1.0.14 gmodels_2.18.1
## [13] xts_0.11-2 colorspace_1.4-1
## [15] rvest_0.3.2 rappdirs_0.3.1
## [17] ggrepel_0.8.0 hoardr_0.5.2
## [19] pan_1.6 xfun_0.5
## [21] crayon_1.3.4 jsonlite_1.6
## [23] hexbin_1.27.2 lme4_1.1-21
## [25] survival_2.43-3 zoo_1.8-5
## [27] iterators_1.0.10 glue_1.3.1
## [29] gtable_0.3.0 webshot_0.5.1
## [31] quantmod_0.4-14 jomo_2.6-7
## [33] scales_1.0.0 mvtnorm_1.0-10
## [35] DBI_1.0.0 miniUI_0.1.1.1
## [37] Rcpp_1.0.1 viridisLite_0.3.0
## [39] units_0.6-2 flashClust_1.01-2
## [41] foreign_0.8-71 httr_1.4.0
## [43] htmlwidgets_1.3 mice_3.4.0
## [45] pkgconfig_2.0.2 XML_3.98-1.19
## [47] nnet_7.3-12 deldir_0.1-16
## [49] crul_0.7.0 labeling_0.3
## [51] tidyselect_0.2.5 rlang_0.3.2
## [53] manipulateWidget_0.10.0 later_0.8.0
## [55] munsell_0.5.0 tools_3.5.3
## [57] generics_0.0.2 broom_0.5.1
## [59] evaluate_0.13 yaml_2.2.0
## [61] purrr_0.3.2 mitml_0.3-7
## [63] nlme_3.1-137 mime_0.6
## [65] leaps_3.0 xml2_1.2.0
## [67] rstudioapi_0.10 compiler_3.5.3
## [69] curl_3.3 e1071_1.7-1
## [71] tibble_2.1.1 stringi_1.4.3
## [73] highr_0.8 classInt_0.3-1
## [75] nloptr_1.2.1 pillar_1.3.1
## [77] LearnBayes_2.15.1 httpuv_1.5.0
## [79] R6_2.4.0 promises_1.0.1
## [81] codetools_0.2-16 boot_1.3-20
## [83] gtools_3.8.1 assertthat_0.2.1
## [85] withr_2.1.2 httpcode_0.2.0
## [87] hms_0.4.2 expm_0.999-4
## [89] parallel_3.5.3 quadprog_1.5-5
## [91] rpart_4.1-13 tidyr_0.8.3
## [93] coda_0.19-2 class_7.3-15
## [95] minqa_1.2.4 rmarkdown_1.12
## [97] TTR_0.23-4 scatterplot3d_0.3-41
## [99] shiny_1.2.0