In arthurgailes/rsegregation: Calculate Empirical Measures of Segregation

Trial and error tests with log base (groups) vs natural log.

devtools::load_all()

Test the difference between total and black/white segregation using ln and base(groups)

# test the difff between ln and log(groups)

detroit_mod <- detroit_race %>%
  mutate(newrace=NA_real_, white_bw = white/(white+black), black_bw=black/(white+black),
    pop_bw = population*(white+black)) %>%
  summarize(
    divln = divergence(across(black:hispanic), population=population, summed=T),
    divln1 = divergence(white_bw, black_bw, population=pop_bw, summed=T),
    divbase = divergence(across(black:hispanic), population=population, logBase=4, summed=T),
    divbase1 = divergence(white_bw, black_bw, newrace, population=pop_bw, logBase=2, summed=T))
detroit_mod

t <- detroit_race %>%
  select(-tract) %>%
  mutate(place_name = ifelse(grepl('Detroit',place_name), place_name,'Other'))
decompose_divergence(t, groupCol='place_name',popCol='population', output='all')

decompose_divergence(t, groupCol='place_name',popCol='population', output='all', logBase=6)

Test whether adding a new group that is completely uniform changes the scores

detroit_new <- detroit_race %>% 
  mutate(across(black:nhpi, ~(.x*.8),), newRace = 0.2) %>% 
  summarize(
    divln = divergence(across(black:nhpi), population=population, summed=T),
    divln1 = divergence(across(black:newRace), population=population, summed=T),
    divbase = divergence(across(black:nhpi), population=population, logBase=6, summed=T),
    divbase1 = divergence(across(black:newRace),  population=population, logBase=7, summed=T))

detroit_new

Intuition

Using ln means that adding a completely desgregated group does nothing to the score - the total amount of measured segregation is still there. Using base (group) diminishes the segregation - the average segregation has gone down.

arthurgailes/rsegregation documentation built on May 23, 2021, 6:33 a.m.