Day 3: Data visualization - Part 2

BSTA 511/611 Fall 2024 Day 5, OHSU

Meike Niederhausen, PhD

2024-10-16

Back to research question

Case study: discrimination in developmental disability support (1.7.1)

  • Previous research
    • Researchers examined DDS expenditures for developmentally disabled residents by ethnicity
    • Found that the mean annual expenditures on Hispanics was less than that on White non-Hispanics.
  • Result: an allegation of ethnic discrimination was brought against the California DDS.
  • Question: Are the data sufficient evidence of ethnic discrimination?

Load dds.discr dataset from oibiostat package

  • The textbook’s datasets are in the R package oibiostat

  • Make sure the oibiostat package is installed before running the code below.

  • Load the oibiostat package and the dataset dds.discr

the code below needs to be run every time you restart R or render a Qmd file

library(oibiostat)
data("dds.discr")
  • After loading the dataset dds.discr using data("dds.discr"), you will see dds.discr in the Data list of the Environment window.

glimpse()

New: glimpse()

  • Use glimpse() from the tidyverse package (technically it’s from the dplyr package) to get information about variable types.
  • glimpse() tends to have nicer output for tibbles than str()
library(tidyverse)
glimpse(dds.discr)  # from tidyverse package (dplyr)
Rows: 1,000
Columns: 6
$ id           <int> 10210, 10409, 10486, 10538, 10568, 10690, 10711, 10778, 1…
$ age.cohort   <fct> 13-17, 22-50, 0-5, 18-21, 13-17, 13-17, 13-17, 13-17, 13-…
$ age          <int> 17, 37, 3, 19, 13, 15, 13, 17, 14, 13, 13, 14, 15, 17, 20…
$ gender       <fct> Female, Male, Male, Female, Male, Female, Female, Male, F…
$ expenditures <int> 2113, 41924, 1454, 6400, 4412, 4566, 3915, 3873, 5021, 28…
$ ethnicity    <fct> White not Hispanic, White not Hispanic, Hispanic, Hispani…

Recall previous data viz

Visualize in more detail:

ethnicity, age, and expenditures (code on next slide)

Code for Visualize in more detail: ethnicity, age, and expenditures

Plot on previous slide

dds.discr_Hips_WhnH <- dds.discr %>%  
  filter(ethnicity == "White not Hispanic" | ethnicity == "Hispanic" ) %>% 
  droplevels()   # remove empty factor levels

ggplot(data = dds.discr_Hips_WhnH,   
       aes(x = expenditures,
           y = age.cohort)) + 
  geom_boxplot(color="darkgrey") + 
  facet_grid(rows = "ethnicity") +   
  labs(x = "Annual Expenditures ($)",
       y = "Race and ethnicity") +
  geom_jitter(     
    aes(color = ethnicity),      
    alpha = 0.3,      
    show.legend = FALSE,      
    position = position_jitter(     
      height = 0.4))

Mean annual DDS expenditures by race/ethnicity: default long format

mean_expend <- 
  dds.discr_Hips_WhnH %>% 
  group_by(
    ethnicity, age.cohort)%>% 
  summarize(
    ave = mean(expenditures))
mean_expend
# A tibble: 12 × 3
# Groups:   ethnicity [2]
   ethnicity          age.cohort    ave
   <fct>              <fct>       <dbl>
 1 Hispanic           0-5         1393.
 2 Hispanic           6-12        2312.
 3 Hispanic           13-17       3955.
 4 Hispanic           18-21       9960.
 5 Hispanic           22-50      40924.
 6 Hispanic           51+        55585 
 7 White not Hispanic 0-5         1367.
 8 White not Hispanic 6-12        2052.
 9 White not Hispanic 13-17       3904.
10 White not Hispanic 18-21      10133.
11 White not Hispanic 22-50      40188.
12 White not Hispanic 51+        52670.

Mean annual DDS expenditures by race/ethnicity: wide format

mean_expend_wide <- 
  mean_expend %>% 
  pivot_wider(
    names_from = ethnicity,
    values_from = ave)
mean_expend_wide
# A tibble: 6 × 3
  age.cohort Hispanic `White not Hispanic`
  <fct>         <dbl>                <dbl>
1 0-5           1393.                1367.
2 6-12          2312.                2052.
3 13-17         3955.                3904.
4 18-21         9960.               10133.
5 22-50        40924.               40188.
6 51+          55585                52670.

Differences in mean annual DDS expenditures by age cohort and race/ethnicity

mean_expend_wide <- mean_expend_wide %>% 
  mutate(diff_mean = `White not Hispanic` - Hispanic)

mean_expend_wide
# A tibble: 6 × 4
  age.cohort Hispanic `White not Hispanic` diff_mean
  <fct>         <dbl>                <dbl>     <dbl>
1 0-5           1393.                1367.     -26.3
2 6-12          2312.                2052.    -260. 
3 13-17         3955.                3904.     -50.9
4 18-21         9960.               10133.     173. 
5 22-50        40924.               40188.    -736. 
6 51+          55585                52670.   -2915.

Question: Are the data sufficient evidence of ethnic discrimination in DDS expenditures when comparing Hispanics with White non-Hispanics?

Simpson’s paradox

  • This case study is an example of confounding known as Simpson’s paradox

  • Simpson’s paradox happens when an association observed in several groups disappears or reverses direction when the groups are combined.

  • In other words, an association between two variables \(X\) and \(Y\) may disappear or reverse direction once data are partitioned into subpopulations based on a third variable \(Z\) (i.e., a confounding variable).

The tidyverse

Artwork by @allison_horst

Tools for wrangling data

  • tidyverse functions
    • tidyverse is a suite of packages that implement tidy methods for data importing, cleaning, wrangling, and visualizing
    • load the tidyverse packages by running the code library(tidyverse)
      • Don’t forget to first install tidyverse!
  • Functions to easily work with rows and columns, such as
    • subset rows/columns
    • add new rows/columns
    • join together different data sets
    • make data long or wide
  • There are often many steps to tidy data
    • we string together commands
    • to be performed sequentially
    • using pipes %>%

Summary of data wrangling so far

  • The pipe %>% to string together commands in sequence
  • mutate() to add a new variable to a dataset
  • select() to select columns (or deselect columns with -variable)
  • filter() to select specific rows
  • pivot_wider() to reshape a dataset from a long to a wide format

Summarizing data

  • tabyl() from janitor package to make frequency tables of categorical variables
  • summarize() to get summary statistics of variables
  • group_by() to group data by categorical variables before finding summaries

What packages are included in the tidyverse?

Core packages
These automatically load when loading the tidyverse package

https://www.tidyverse.org/

List of all packages:

tidyverse_packages(include_self = TRUE)
 [1] "broom"         "conflicted"    "cli"           "dbplyr"       
 [5] "dplyr"         "dtplyr"        "forcats"       "ggplot2"      
 [9] "googledrive"   "googlesheets4" "haven"         "hms"          
[13] "httr"          "jsonlite"      "lubridate"     "magrittr"     
[17] "modelr"        "pillar"        "purrr"         "ragg"         
[21] "readr"         "readxl"        "reprex"        "rlang"        
[25] "rstudioapi"    "rvest"         "stringr"       "tibble"       
[29] "tidyr"         "xml2"          "tidyverse"