get_coverage.R

library(rstan)
library(data.table)
library(tidyverse)
library(RColorBrewer)
library(ggplot2)
library(cowplot)
library(foreach)

TIME_CORRECTION_TYPE <<- 'beta'
NCPU <<- 2
PREPROCESS_DATA <<-  TRUE

source('config/config.R')
source(paste0(PROJECT_PATH, '/config/file_paths.R'))
source(paste0(PROJECT_PATH, '/scripts/model_fitting_functions.R'))
source(paste0(PROJECT_PATH, '/plotting_scripts/plotting_functions.R'))
source(paste0(PROJECT_PATH, '/scripts/model_prediction_evaluation_functions.R'))
source(paste0(PROJECT_PATH, '/scripts/model_cross_validation_functions.R'))

infant_data = configure_data(TRAINING_INFANT_DATA_PATH)
data = infant_data
data$number = seq(1, length(data$is_post))
data = as.data.table(data)

name = str_split(TRAINING_INFANT_DATA_PATH, '/')[[1]][length(str_split(TRAINING_INFANT_DATA_PATH, '/')[[1]])]
name = str_replace(name, '.csv', '')
file_name = file.path(OUTPUT_PATH, 'model_loocv', paste0('loocv_posteriors_', name, '_', TIME_CORRECTION_TYPE, '.csv'))
loocv = fread(file_name)

results = get_posterior_prediction_coverage(loocv, data)
print(results)

# get accuracy
intervals = results$data
intervals[, diff := q0.945 - q0.255]
intervals[, median(diff), by = fragment]