poststratify() should return posterior samples

[jamesdunham]

The current behavior is to return poststratified means. In typical use, we want poststratified means and CIs. To make this straightforward, first allow returning poststratified posterior samples. Then summarize the poststratified posterior.

[jamesdunham]

Basic workaround:

library(dgo)
library(data.table)

data(toy_dgirtfit)
d = as.data.frame(toy_dgirtfit)
stopifnot(is.data.table(d))

samples = lapply(unique(d$iteration), function(x) {
  poststratify(d[iteration == x], annual_state_race_targets, 
    strata_names = c('state', 'year'), aggregated_names = 'race3')
})
samples = rbindlist(samples, idcol = 'iteration')

head(samples)
#    iteration state year        value
# 1:         1    SC 2009 -0.252734127
# 2:         1    SC 2010 -0.651696989
# 3:         1    VA 2009 -0.263817901
# 4:         1    VA 2010 -0.006465143
# 5:         2    SC 2009 -0.169913600
# 6:         2    SC 2010 -0.342846234

# The means of these poststratified posterior samples should equal the
# postratified means returned by poststratify()

means = poststratify(toy_dgirtfit, annual_state_race_targets, 
  strata_names = c('state', 'year'), aggregated_names = 'race3')
means
#    state year      value
# 1:    SC 2009 -0.6391259
# 2:    SC 2010 -0.7397078
# 3:    VA 2009 -0.4158158
# 4:    VA 2010 -0.3251222

mean_samples = samples[, .(value = mean(value)), by = c('state', 'year')]
mean_samples 
#    state year      value
# 1:    SC 2009 -0.6391259
# 2:    SC 2010 -0.7397078
# 3:    VA 2009 -0.4158158
# 4:    VA 2010 -0.3251222

stopifnot(all.equal(means, mean_samples, check.attributes = FALSE))