exploring more robust alternatives to https://github.com/az-digitalag/state_organic_ag/pull/5/files#diff-1fb0826bac1b44806443765a86931b5c4b9ed8a35d16d8083f80906ef8fd7d49

model_output_munging.R

sum_out <- broom.mixed::tidyMCMC(coda_out, conf.int = TRUE, conf.level = 0.95) %>% 
  rename(mean = estimate, sd = std.error, pc2.5 = conf.low, pc97.5 = conf.high) %>% 
  mutate(sig = if_else(pc2.5 * pc97.5 > 0, TRUE, FALSE)) # test if UCL and LCL both have the same sign

years <- data.frame(year_index = 1:length(unique(x$year)), year = unique(x$year))
states <- data.frame(state_index = 1:length(unique(x$state)), state = unique(x$state))

sum_rep <- broom.mixed::tidyMCMC(coda_rep, conf.int = TRUE, conf.level = 0.95) %>% 
  rename(mean = estimate, sd = std.error, pc2.5 = conf.low, pc97.5 = conf.high) %>% 
  separate(term, sep = '\\[|\\]|,', into = c('term', 'i', 'j', 'other'), convert = TRUE) %>% 
  mutate(var = case_when(j == 1 ~ "area~(kha)",
                         j == 2 ~ "number~(k)",
                         j == 3 ~ "sales~(MM)"),
         year_index = i %% 6 + 1,
         state_index = i %% 50 + 1) %>% 
  left_join(years) %>% 
  left_join(states)

sum_rep2 <- postjags::coda.fast(chains = 3, burn.in = 0, thin = 1, coda = coda_rep)
sum_rep$year <- rep(dat$year + 2014, each = 3)
sum_rep$Var <- rep(c("area~(kha)", "number~(k)", "sales~(MM)"), 300)
sum_rep$state <- rep(dat$state, each = 3)