Output of Total (Knowns + Unknowns)

Method:

library(dplyr)
library(Syndemics)
library(doParallel)

results <- list()
unique_years <- unique(data$year)
n_cores <- detectCores() - 1

registerDoParallel(n_cores)
for(i in unique_years) {
  estimates <- foreach(j = 1:1000, .combine = 'rbind', .packages = c("dplyr", "Syndemics")) %dopar% {
    set.seed(j)
    
    output <- data %>% 
      mutate(N_ID = ifelse(N_ID == -1, round(runif(nrow(.), 1, 10)), N_ID)) %>%
      filter(year == i) %>% 
      select(-c(final_re, final_sex, age_grp_ten)) %>%
      unique()
    
    Syndemics::crc(output, "N_ID", c("Casemix", "APCD", "BSAS", "PMP", "Matris", "Death"),
                   formula.selection = "stepwise", method = "poisson", 
                   opts.stepwise = list(direction = "forward",
                                        verbose = FALSE))$estimate
  }
  results[[as.character(i)]] <- list(crcMean = mean(estimates), crcMinimum = min(estimates), crcMaximum = max(estimates))
}
stopImplicitCluster()

output <- do.call(rbind, results) %>% 
  as.data.frame() %>%
  tibble::rownames_to_column(var = "year") %>%
  mutate(year = as.integer(year)) %>%
  left_join(., data %>% 
              select(-c(final_re, final_sex, age_grp_ten)) %>%
              unique() %>%
              group_by(year) %>%
              summarise(., known = sum(N_ID))
            ) %>% 
  rowwise() %>%
  mutate(estimate = crcMean + known,
         estimateMin = crcMinimum + known,
         estimateMax = crcMaximum + known)

First Implementation of CRC on PHD OUDCounts

What's Changed

• AIC Poisson/Neg-Bin Implementation by @ryan-odea in #1
• Removed good-turing, fixed some bugs with formula_list by @ryan-odea in #2
• Stepwise by @ryan-odea in #3
• Stepwise by @ryan-odea in #4

New Contributors

• @ryan-odea made their first contribution in #1

Full Changelog: https://github.com/SyndemicsLab/Syndemics/commits/v1.0.0

CRC Code:

library(dplyr)
library(Syndemics)
library(doParallel)

results <- list()
unique_years <- unique(data$year)

for(i in unique_years) {
  estimates <- foreach(j = 1:1000, .combine = 'c', .packages = c("dplyr", "Syndemics")) %dopar% {
    set.seed(j)
    
    output <- data %>% 
      mutate(N_ID = ifelse(N_ID == -1, round(runif(nrow(.), 1, 10)), N_ID)) %>%
      filter(year == i)
    
    Syndemics::crc(output, "N_ID", c("Casemix", "APCD", "BSAS", "PMP", "Matris", "Death"),
                   formula.selection = "stepwise", method = "poisson", 
                   opts.stepwise = list(direction = "forward",
                                        verbose = FALSE))$estimate
  }
  results[[as.character(i)]] <- mean(estimates)
}

stopCluster()
print(results)