Using data from CHR 2018 (and 2013)
An HTML version of this document is available to view at https://rpubs.com/TELOVE/ProjectA-sample-plan-431-2024.
A template for the Project A Plan is available to you, via the Examples page on the Project A website. Please use it in combination with this document to prepare revisions, as needed, to your Plan. We used it to develop this document.
You need to write your own comments and responses to the Plan’s requirements. You are welcome to use the words here as illustrative examples of what we’re looking for, but these should be edited by you to be specific to your project.
You need a real title (80 characters, maximum, without using “431” or “Project” or “Project A”) in your Plan. You can, as we have above, include a subtitle, but the main title must stand on its own. Of course, in this sample plan, we used some words you’re not allowed to use, and we will break other rules (and note them) in what follows.
knitr::opts_chunk$set(comment = NA)
library(janitor)
library(knitr)
library(naniar)
library(xfun)
library(easystats)
library(tidyverse)
theme_set(theme_bw())
url_script <-
"https://raw.githubusercontent.com/THOMASELOVE/431-data/refs/heads/main/data/Love-431.R"
source(url_script)These are a few of the things we did differently to get our data.
We are pulling data from 2018 here (and 2013 later) through its URL at the County Health Rankings data and documentation site. You are working with 2024 and 2019 data.
In 2013 and 2018, what is now called county_clustered was called county_ranked, and we need to account for that here.
data_2018_url <-
"https://www.countyhealthrankings.org/sites/default/files/analytic_data2018_0.csv"
chr_2018_raw <- read_csv(data_2018_url, skip = 1, guess_max = 4000,
show_col_types = FALSE) |>
rename(county_clustered = county_ranked) |>
select(fipscode, county, state, county_clustered, year,
ends_with("rawvalue")) Next, we filter these data to the rows which have county_clustered values of 1.
chr_2018_raw <- chr_2018_raw |>
filter(county_clustered == 1)The resulting chr_2018_raw tibble now has 3078 rows, and 107 columns.
Make sure you look at the Quarto file for this document, and note the use of inline coding to get R to tell me the number of rows and number of columns in the resulting chr_2018_raw tibble.
Another approach would have been to use the dim() function here.
In selecting six states for this sample plan, we’re using some states you’re not permitted to use. Specifically, we have arbitrarily decided to use New York, Ohio, Massachusetts, Pennsylvania, Maine and North Carolina.
Here, we’ll select our six states, then change the state to a factor variable.
chr_2018 <- chr_2018_raw |>
filter(state %in% c("NY", "OH", "MA", "PA", "ME", "NC")) |>
mutate(state = factor(state))Next, we’ll look to see how many counties are in each state.
chr_2018 |> count(state) # A tibble: 6 × 2
state n
<fct> <int>
1 MA 14
2 ME 16
3 NC 100
4 NY 62
5 OH 88
6 PA 67We have selected 6 states, yielding a total of 347 clustered counties, which is between 300 and 800 so we’re all set.
Again, in this last sentence, we’ve used inline coding to get R to tell me the number of states and the number of rows in the resulting chr_2018 tibble.
Here is the place to put a brief description as to why you selected the states that you selected. we will leave that work to you. As for our reason, these are six states in which Dr. Love has spent pleasant summer vacations.
we have selected a set of five variables for this sample plan. None of these variables were available for you to choose.
We’ve decided to select variables v128, v065, v024, v052 and v122.
chr_2018 <- chr_2018 |>
select(fipscode, state, county, county_clustered,
v128_rawvalue, v065_rawvalue, v024_rawvalue,
v052_rawvalue, v122_rawvalue)we now have a chr_2018 tibble with exactly 9 columns, as required.
The variables we are using describe the following measures:
| Initial Name | New Name | Role | Description | Gathered |
|---|---|---|---|---|
v128_rawvalue |
child_mort |
A1 outcome | Child mortality (deaths among residents under age 18 per 100,000 population) | 2013-16 |
v065_rawvalue |
free_lunch |
A1 predictor | % of children enrolled in public schools that are eligible for free or reduced price lunch | 2015-16 |
v024_rawvalue |
child_pov |
A2 outcome | % of people under 18 in poverty | 2016 |
v052_rawvalue |
below_18 |
A2 predictor | % of county residents below 18 years of age | 2016 |
v122_rawvalue |
unins_kids_2018 |
A3 outcome | % of children under age 19 without health insurance | 2015 |
v065, v024, v052 and v122 are all proportions, that need to be multiplied by 100v128 is OK as isHere, we’ll multiply the four variables that describe proportions by 100 to obtain percentages instead, to ease interpretation.
chr_2018 <- chr_2018 |>
mutate(free_lunch = 100*v065_rawvalue,
child_pov = 100*v024_rawvalue,
below_18 = 100*v052_rawvalue,
unins_kids_2018 = 100*v122_rawvalue,
.keep = "unused") |>
rename(child_mort = v128_rawvalue)dim(chr_2018)[1] 347 9names(chr_2018)[1] "fipscode" "state" "county" "county_clustered"
[5] "child_mort" "free_lunch" "child_pov" "below_18"
[9] "unins_kids_2018" What does this indicate to you about the use of .keep = "unused" in the mutate() function?
.keep = "unused" in mutate() retains only the columns not used in the process of creating new columns. This is useful if, as in this case, you want to generate new columns, but no longer need the columns used to generate them. See this reference on mutate().we renamed v122 as unins_kids_2018 since it was reported in CHR 2018. Soon, we will create unins_kids_2013 as well, for comparison in Analysis 3.
To establish our cut points, we should look at the 40th and 60th percentiles of the existing data for our planned predictor for Analysis 2, which is below_18.
chr_2018 |>
summarise(q40 = quantile(below_18, c(0.4)),
q60 = quantile(below_18, c(0.6)))# A tibble: 1 × 2
q40 q60
<dbl> <dbl>
1 20.4 21.7So we will create a three-level variable where values of 20.4 and lower will fall in the “Low” group, and values of 21.7 and higher will fall in the “High” group1.
chr_2018 <- chr_2018 |>
mutate(below18_grp = case_when(
below_18 <= 20.4 ~ "Low",
below_18 >= 21.7 ~ "High")) |>
mutate(below18_grp = factor(below18_grp))
chr_2018 |> count(below18_grp)# A tibble: 3 × 2
below18_grp n
<fct> <int>
1 High 139
2 Low 139
3 <NA> 69It appears that we have 139 subjects (40% of the original 347) in the High group and the same number in the Low group, with the rest now listed as missing, and the below18_grp variable is now a factor, so that’s fine. (If you have a slightly different number in “High” than in “Low”, that would also be OK, so long as it’s close to 40% in each group.)
In our case, we’ll add data from CHR 2013, since that’s five years prior to the 2018 County Health Rankings report.
Rather than pull this from a .csv file, we will pull it directly from the CHR website, as follows.
The variables we need in our chr_2013_raw file are just the fipscode and our analysis 3 outcome, which starts as the v122_rawvalue variable.
v122_rawvalue) come from Small Area Health Insurance Estimates for 2010. We’ll need that year when we build the codebook, later.data_2013_url <-
"https://www.countyhealthrankings.org/sites/default/files/analytic_data2013.csv"
chr_2013 <- read_csv(data_2013_url, skip = 1, guess_max = 4000,
show_col_types = FALSE) |>
rename(county_clustered = county_ranked) |>
filter(county_clustered == 1) |>
select(fipscode, v122_rawvalue) |>
mutate(unins_kids_2013 = 100*v122_rawvalue,
.keep = "unused")
names(chr_2013)[1] "fipscode" "unins_kids_2013"Now, we’ll join the two files.
chr_2018 <- left_join(chr_2018, chr_2013, by = "fipscode")Finally, we’ll check to see if this has created any missing values (which would happen if a county in CHR 2018 had data on this variable but did not in CHR 2013.)
n_miss(chr_2018$unins_kids_2013)[1] 0Now we arrange the variables in the specified order from Data Task 5, and then save the new result to a new .Rds file.
chr_2018 <- chr_2018 |>
select(fipscode, state, county,
child_mort, free_lunch, ## Analysis 1 variables
child_pov, below18_grp, ## Analysis 2 variables
below_18, ## Quantitative version of group
unins_kids_2018, unins_kids_2013, ## Analysis 3 variables
county_clustered)
write_rds(chr_2018, file = "chr_2018_Thomas_Love.Rds")chr_2018# A tibble: 347 × 11
fipscode state county child_mort free_lunch child_pov below18_grp below_18
<chr> <fct> <chr> <dbl> <dbl> <dbl> <fct> <dbl>
1 23001 ME Androsco… 66.8 56.4 19.5 High 21.8
2 23003 ME Aroostoo… 75.5 53.5 22.1 Low 18.4
3 23005 ME Cumberla… 42.6 32.9 11.3 Low 19.2
4 23007 ME Franklin… 53.8 54.9 20.4 Low 18.1
5 23009 ME Hancock … 49.6 42.8 15.3 Low 17.7
6 23011 ME Kennebec… 48.1 50.5 18.2 Low 19.7
7 23013 ME Knox Cou… 48.0 42.3 16.8 Low 18.1
8 23015 ME Lincoln … 46.3 49.5 19.2 Low 17.2
9 23017 ME Oxford C… 35.9 62.4 20.2 Low 19.0
10 23019 ME Penobsco… 59.8 47.2 18.2 Low 18.3
# ℹ 337 more rows
# ℹ 3 more variables: unins_kids_2018 <dbl>, unins_kids_2013 <dbl>,
# county_clustered <dbl>chr_2018 |> tabyl(state, below18_grp) |>
adorn_totals(where = c("row", "col")) state High Low NA_ Total
MA 2 7 5 14
ME 1 15 0 16
NC 42 35 23 100
NY 15 30 17 62
OH 63 9 16 88
PA 16 43 8 67
Total 139 139 69 347describe_distribution() resultsdescribe_distribution(chr_2018)Variable | Mean | SD | IQR | Range | Skewness | Kurtosis | n | n_Missing
--------------------------------------------------------------------------------------------------
child_mort | 53.66 | 16.59 | 18.09 | [19.96, 125.42] | 1.10 | 2.28 | 309 | 38
free_lunch | 52.29 | 16.61 | 16.51 | [14.93, 99.10] | 0.94 | 1.19 | 333 | 14
child_pov | 21.14 | 7.37 | 9.20 | [4.70, 57.10] | 0.64 | 1.31 | 347 | 0
below_18 | 20.92 | 2.70 | 3.38 | [5.15, 32.29] | -0.54 | 3.96 | 347 | 0
unins_kids_2018 | 4.70 | 1.73 | 1.96 | [0.83, 17.96] | 1.44 | 9.49 | 347 | 0
unins_kids_2013 | 6.80 | 2.12 | 2.60 | [1.50, 16.80] | 0.51 | 2.08 | 347 | 0
county_clustered | 1.00 | 0.00 | 0.00 | [1.00, 1.00] | | | 347 | 0Here, we have minimum and maximum values that make sense for all of the quantitative variables in our data. All of the data reflect information for the appropriate number of counties, since, as mentioned previously, we have missing values in the two Analysis 1 variables we selected (child_mort and free_lunch.)
data_codebook() resultsdata_codebook(chr_2018, max_values = 6, range_at = 15)chr_2018 (347 rows and 11 variables, 11 shown)
ID | Name | Type | Missings | Values | N
---+------------------+-------------+------------+------------------+-------------
1 | fipscode | character | 0 (0.0%) | 23001 | 1 ( 0.3%)
| | | | 23003 | 1 ( 0.3%)
| | | | 23005 | 1 ( 0.3%)
| | | | 23007 | 1 ( 0.3%)
| | | | 23009 | 1 ( 0.3%)
| | | | 23011 | 1 ( 0.3%)
| | | | (...) |
---+------------------+-------------+------------+------------------+-------------
2 | state | categorical | 0 (0.0%) | MA | 14 ( 4.0%)
| | | | ME | 16 ( 4.6%)
| | | | NC | 100 ( 28.8%)
| | | | NY | 62 ( 17.9%)
| | | | OH | 88 ( 25.4%)
| | | | PA | 67 ( 19.3%)
---+------------------+-------------+------------+------------------+-------------
3 | county | character | 0 (0.0%) | Adams County | 2 ( 0.6%)
| | | | Alamance County | 1 ( 0.3%)
| | | | Albany County | 1 ( 0.3%)
| | | | Alexander County | 1 ( 0.3%)
| | | | Allegany County | 1 ( 0.3%)
| | | | Alleghany County | 1 ( 0.3%)
| | | | (...) |
---+------------------+-------------+------------+------------------+-------------
4 | child_mort | numeric | 38 (11.0%) | [19.96, 125.42] | 309
---+------------------+-------------+------------+------------------+-------------
5 | free_lunch | numeric | 14 (4.0%) | [14.93, 99.1] | 333
---+------------------+-------------+------------+------------------+-------------
6 | child_pov | numeric | 0 (0.0%) | [4.7, 57.1] | 347
---+------------------+-------------+------------+------------------+-------------
7 | below18_grp | categorical | 69 (19.9%) | High | 139 ( 50.0%)
| | | | Low | 139 ( 50.0%)
---+------------------+-------------+------------+------------------+-------------
8 | below_18 | numeric | 0 (0.0%) | [5.15, 32.29] | 347
---+------------------+-------------+------------+------------------+-------------
9 | unins_kids_2018 | numeric | 0 (0.0%) | [0.83, 17.96] | 347
---+------------------+-------------+------------+------------------+-------------
10 | unins_kids_2013 | numeric | 0 (0.0%) | [1.5, 16.8] | 347
---+------------------+-------------+------------+------------------+-------------
11 | county_clustered | numeric | 0 (0.0%) | 1 | 347 (100.0%)
----------------------------------------------------------------------------------So, we pass all of the necessary checks.
What I originally told you to do was this:
chr_2018 |>
summarise(across(everything(), ~ n_distinct(.)))# A tibble: 1 × 11
fipscode state county child_mort free_lunch child_pov below18_grp below_18
<int> <int> <int> <int> <int> <int> <int> <int>
1 347 6 282 310 334 194 3 347
# ℹ 3 more variables: unins_kids_2018 <int>, unins_kids_2013 <int>,
# county_clustered <int>but the problem here is that some of the results we want to see don’t turn up in the printed output.
You could show, for instance, the counts for the last five variables in another call to this function, as follows:
chr_2018 |>
summarise(across(everything(), ~ n_distinct(.))) |>
select(7:11)# A tibble: 1 × 5
below18_grp below_18 unins_kids_2018 unins_kids_2013 county_clustered
<int> <int> <int> <int> <int>
1 3 347 346 84 1Another simple and attractive enough way to show the results of this check for all 11 variables, is to use the kable() function from the knitr package, as we have done below. There we can see all eleven results if we scroll through the HTML.
chr_2018 |>
summarise(across(everything(), ~ n_distinct(.))) |>
kable()| fipscode | state | county | child_mort | free_lunch | child_pov | below18_grp | below_18 | unins_kids_2018 | unins_kids_2013 | county_clustered |
|---|---|---|---|---|---|---|---|---|---|---|
| 347 | 6 | 282 | 310 | 334 | 194 | 3 | 347 | 346 | 84 | 1 |
child_mort, child_pov and unins_kids in each year) and in our quantitative predictor (free_lunch) for Analysis 1.county_clustered for every row of our data, so that’s correct, too.So we pass all of the necessary checks here, as well.
Here’s a way to avoid the scrolling window in HTML…
tab10_4 <- chr_2018 |>
summarise(across(everything(), ~ n_distinct(.)))
tab10_4 |> select(1:5) |> kable()| fipscode | state | county | child_mort | free_lunch |
|---|---|---|---|---|
| 347 | 6 | 282 | 310 | 334 |
tab10_4 |> select(6:11) |> kable()| child_pov | below18_grp | below_18 | unins_kids_2018 | unins_kids_2013 | county_clustered |
|---|---|---|---|---|---|
| 194 | 3 | 347 | 346 | 84 | 1 |
There are other, fancier, approaches we could use, but we will be happy with any of these, so long as we can see the results for all 11 columns.
Our chr_2018 tibble contains 347 counties and 11 variables.
| Variable | Role | Old Name | Description | Year(s) |
|---|---|---|---|---|
| fipscode | ID | fipscode |
FIPS code | – |
| state | ID | state |
State Abbreviation (OH, MA, ME, NC, NY, PA) | – |
| county | ID | county |
County Name | – |
| child_mort | A1 outcome | v128 |
Child mortality (deaths among residents under age 18 per 100,000 population) | 2013-16 |
| free_lunch | A1 predictor | v065 |
% of children enrolled in public schools that are eligible for free or reduced price lunch | 2015-16 |
| child_pov | A2 outcome | v024 |
% of people under 18 in poverty | 2016 |
| below_18_grp | A2 predictor | - | Low (below_18 \(\leq\) 20.4) or High (below_18 \(\geq\) 21.7) % of county residents below 18 years of age | 2016 |
| below_18 | – | v052 |
% of county residents below 18 years of age | 2016 |
| unins_kids_2018 | A3 outcome | v122 |
% of children under age 19 without health insurance, CHR 2018 | 2015 |
| unins_kids_2013 | A3 outcome | v122 |
% of children under age 19 without health insurance, CHR 2013 | 2010 |
| county_clustered | - | county_clustered |
Indicates county is ranked (all values are 1, as required) | 2024 |
Here is where you’ll place your research question for Analysis 1, which in our case involves predicting child_mort from free_lunch.
Here is where you’ll place your research question for Analysis 2, which in our case involves comparing means of child_pov across our two groups in below18_grp.
Here is where you’ll place your research question for Analysis 3, which in our case involves comparing means of unins_kids in the 2018 report (where the data were measured in 2015) as compared to the 2013 report by CHR (where the data come from 2010).
Here is where you’ll place your reflection. We’ll leave that to you.
xfun::session_info()R version 4.4.1 (2024-06-14 ucrt)
Platform: x86_64-w64-mingw32/x64
Running under: Windows 11 x64 (build 22631)
Locale:
LC_COLLATE=English_United States.utf8
LC_CTYPE=English_United States.utf8
LC_MONETARY=English_United States.utf8
LC_NUMERIC=C
LC_TIME=English_United States.utf8
Package version:
askpass_1.2.0 backports_1.5.0 base64enc_0.1.3
bayestestR_0.14.0 bit_4.0.5 bit64_4.0.5
blob_1.2.4 broom_1.0.6 bslib_0.8.0
cachem_1.1.0 callr_3.7.6 cellranger_1.1.0
cli_3.6.3 clipr_0.8.0 coda_0.19-4.1
codetools_0.2-20 colorspace_2.1-1 compiler_4.4.1
conflicted_1.2.0 correlation_0.8.5 cpp11_0.5.0
crayon_1.5.3 curl_5.2.2 data.table_1.16.0
datasets_4.4.1 datawizard_0.12.3 DBI_1.2.3
dbplyr_2.5.0 digest_0.6.37 dplyr_1.1.4
dtplyr_1.3.1 easystats_0.7.3 effectsize_0.8.9
emmeans_1.10.4 estimability_1.5.1 evaluate_1.0.0
fansi_1.0.6 farver_2.1.2 fastmap_1.2.0
fontawesome_0.5.2 forcats_1.0.0 fs_1.6.4
gargle_1.5.2 generics_0.1.3 ggplot2_3.5.1
glue_1.7.0 googledrive_2.1.1 googlesheets4_1.1.1
graphics_4.4.1 grDevices_4.4.1 grid_4.4.1
gridExtra_2.3 gtable_0.3.5 haven_2.5.4
highr_0.11 hms_1.1.3 htmltools_0.5.8.1
htmlwidgets_1.6.4 httr_1.4.7 ids_1.0.1
insight_0.20.4 isoband_0.2.7 janitor_2.2.0
jquerylib_0.1.4 jsonlite_1.8.9 knitr_1.48
labeling_0.4.3 lattice_0.22-6 lifecycle_1.0.4
lubridate_1.9.3 magrittr_2.0.3 MASS_7.3-61
Matrix_1.7-0 memoise_2.0.1 methods_4.4.1
mgcv_1.9.1 mime_0.12 modelbased_0.8.8
modelr_0.1.11 multcomp_1.4-26 munsell_0.5.1
mvtnorm_1.3-1 naniar_1.1.0 nlme_3.1.164
norm_1.0.11.1 numDeriv_2016.8.1.1 openssl_2.2.1
parallel_4.4.1 parameters_0.22.2 performance_0.12.3
pillar_1.9.0 pkgconfig_2.0.3 plyr_1.8.9
prettyunits_1.2.0 processx_3.8.4 progress_1.2.3
ps_1.8.0 purrr_1.0.2 R6_2.5.1
ragg_1.3.2 rappdirs_0.3.3 RColorBrewer_1.1.3
Rcpp_1.0.13 readr_2.1.5 readxl_1.4.3
rematch_2.0.0 rematch2_2.1.2 report_0.5.9
reprex_2.1.1 rlang_1.1.4 rmarkdown_2.28
rstudioapi_0.16.0 rvest_1.0.4 sandwich_3.1-1
sass_0.4.9 scales_1.3.0 see_0.9.0
selectr_0.4.2 snakecase_0.11.1 splines_4.4.1
stats_4.4.1 stringi_1.8.4 stringr_1.5.1
survival_3.7-0 sys_3.4.2 systemfonts_1.1.0
textshaping_0.4.0 TH.data_1.1-2 tibble_3.2.1
tidyr_1.3.1 tidyselect_1.2.1 tidyverse_2.0.0
timechange_0.3.0 tinytex_0.53 tools_4.4.1
tzdb_0.4.0 UpSetR_1.4.0 utf8_1.2.4
utils_4.4.1 uuid_1.2.1 vctrs_0.6.5
viridis_0.6.5 viridisLite_0.4.2 visdat_0.6.0
vroom_1.6.5 withr_3.0.1 xfun_0.47
xml2_1.3.6 xtable_1.8-4 yaml_2.3.10
zoo_1.8-12 We could have chosen to use “less than 20.4” and “higher than 21.7” as well, which would potentially have a small impact on our final groups.↩︎