covidCensus

The goal of covidCensus is to easily join county level COVID-19 case and death data with other information from the census (eg income, demographic statistics, population density). The purpose of its creation was educational. I had been exploring COVID-19 data for awhile and wanted to observe patterns between health outcomes and other social markers.

Note: I am not a professional researcher, just a hobbyist. Any findings from this study should be taken with a grain of salt.

Installation

covidCensus is not on CRAN, but this is how to install it from Github.

devtools::install_github("shuckle16/covidCensus")

Package Structure

Currently there are 4 fetch_* functions for retrieving the data you want. Each of them returns the raw response from an http request.

• fetch_nyt
• fetch_census_dens
• fetch_census_demo
• fetch_census_income

There are also 4 corresponding tidy_* functions to clean up the above data so they can be joined together. Each of them returns a tidy dataframe (tibble) where each row represents a county.

• tidy_nyt
• tidy_census_dens
• tidy_census_demo
• tidy_census_income

Example

Some sample code which uses the package to plot the correlation between median income and virus deaths per capita.

library(dplyr)
library(ggplot2)
library(covidCensus)

tidied_nyt <- fetch_nyt() %>% tidy_nyt()

tidied_census_dens <- fetch_census_dens() %>% tidy_census_dens()

tidied_census_income <- fetch_census_income() %>% tidy_census_income()

join_datasets <- function() {
  tidied_census_income %>%
    inner_join(tidied_census_dens, by = c("county", "state")) %>% 
    inner_join(tidied_nyt)
}

joined_data <- join_datasets() 

joined_data %>%   
  filter(county != "New York") %>% 
  ggplot(aes(x = median_income, y = (deaths / pop) * 100000)) + 
  geom_point(alpha = 0.7) + 
  scale_x_continuous(labels = scales::comma) +
  xlab("Median Annual Income ($)") + 
  ylab("Deaths per 100k") + 
  ggtitle(
    "COVID-19 Death rate in the USA versus median income", 
    subtitle = paste("Each dot is a county. Deaths data from", max(tidied_nyt$date))
    )

To do list

• Consider using {polite} package instead of xml2. There might be a nicer way to make the census api calls, too. Suggestions welcome
• Find other interesting county-level variables
• Add tests? consistent names from the tidy functions, etc
• Add a vignette or two
• Fix the warnings in tidy_census_income