Hierarchical Bayes Shrinkage — hb

This is a function that takes a "success" and a "total" variable (often something like "number of devices staying home" and "total number of devices") and shrinks them to the full data set using shrinkage methods for proportions.

hb_shrink(success, total)

Arguments

success: A numeric integer variable containing the number of successes.
total: A numeric integer variable containing the total sample size.

Details

This is usually called by group, either with dplyr::group_by or with the by argument in a data.table, so that individual observations can be shrunk to the group level.

Examples

if (FALSE) {
# The directory distdat is the folder we have downloaded the distancing data to from AWS.
# Read and compile all distancing data from May 1 to May 7
distancing <- read_distancing(
    start = lubridate::ymd('2020-05-01'),
    end = lubridate::ymd('2020-05-07'),
    dir = distdat
)

# Shrink county to state
distancing <- distancing[,.(county_fips = county_fips,
                           unshrunk_stay_home = completely_home_device_count/device_count,
                           shrunk_stay_home = hb_shrink(completely_home_device_count, device_count)),
                           by = .(state_fips, date)]

}