R/raw2logit.R

In changxiulee/BayesianRasch: Bayesian Rasch Modeling

#' Convert raw score to logit score
#'
#' This function convert the raw score to logit score using RSM model assuming zero uncertainty
#' with beta and thres parameter
#'
#' @param data A data frame containing the data
#' @param item Item to be converted to logit score
#' @param beta beta parameter of RSM
#' @param thres Andrich threshold
#' @param n.chains Number of chains to be used in MCMC
#' @export raw2logit

raw2logit = function(data, item, beta, thres, n.chains){
  item.variable = data[, item]
  p = ncol(item.variable)
  N  = c(nrow(item.variable))
  group = cbind(rep(1, N), seq(1, N, 1))
  if(min(apply(item.variable, 2, min,na.rm = TRUE)) == 0){
    item.variable = item.variable + 1
  }
  K = max(apply(item.variable, 2, max,na.rm = TRUE))
  N = nrow(item.variable)
  G = 1
  model = runjags::run.jags(model = system.file("JAGS","R2L.txt", package = "BayesianRasch" ), monitor = c("beta", "theta", "thres"),
                   data = list(Y = as.matrix(item.variable), Gr = group, G = G, N = N, K = K, p = p,
                               beta = beta, thres = thres), n.chains = n.chains,
                   burnin = 0, adapt = 1000, sample = round(2100 / 3), method = "rjparallel")
  mcmc = coda::as.mcmc.list(model)
  n.chains = length(mcmc)
  mcmc.matrix = NULL
  for(i in 1:n.chains){mcmc.matrix = rbind(mcmc.matrix, mcmc[[i]][,])}
  mcmc.matrix = BRSM.transform(mcmc.matrix, item, max(K))
  obj = list(mcmc = mcmc.matrix, data = data, item = item, N = N, DIF = FALSE)
  class(obj) = "BRSM"
  obj
}