R/modelcompare.R

In bdlim: Bayesian Distributed Lag Interaction Models

#' Model comparison for bdlim objects
#'
#' @param object An object of class bdlim4 obtained from the bdlim4 function.
#'
#' @return A vector of model probabilities.
#' @export

modelcompare <- function(object) {
  #  iterations to keep.
  iter_keep <- seq(object$nburn + 1, object$nits, by = object$nthin)

  # make model probabilities
  modelselect <- table(apply(object$loglik[iter_keep, ], 1, which.max))
  names(modelselect) <- colnames(object$loglik)[as.numeric(names(modelselect))]

  # add in any groups missing due to 0 probability
  temp <- object$loglik[1, ]
  temp[] <- 0
  temp[names(modelselect)] <- c(modelselect)
  temp <- unlist(temp / sum(temp))

  return(temp)
}