R/selection_criteria.R

In julieaubert/cobiclust: Biclustering via Latent Block Model Adapted to Overdispersed Count Data

# cobiclust R package
# Copyright INRAE 2020
# Universite Paris-Saclay, AgroParisTech, INRAE, UMR MIA-Paris, 75005, Paris, France
####################################################################################
#' Calculate selection criteria.
#'
#' @param x The output of the cobiclust function.
#' @param K The number of groups in rows.
#' @param G The number of groups in columns.
#' @return A dataframe with 7 columns.
#' \describe{
#' \item{\code{vICL}}{the vICL selection criterion.}
#' \item{\code{BIC}}{the BIC selection criterion.}
#' \item{\code{penKG}}{the value of the BIC penalty.}
#' \item{\code{lb}}{the value of the lower bound of the log-likelihood.}
#' \item{\code{entZW}}{the value of the entropy of the latent variables Z and W.}
#' \item{\code{K}}{the number of groups in rows.}
#' \item{\code{G}}{the number of groups in columns.}
#' }
#' @export

selection_criteria <-
  function(x, K, G){
    if (K != x$K) {
      warning('K and x$K are not the same. K will be ignored.')
    }
    if (G != x$G) {
      warning('G and x$G are not the same. G will be ignored.')
    }
    K <- x$K
    G <- x$G
    lb <- x$info$lb
    penKG <- penalty(x)
    BIC <- lb - penKG
    ent_ZW <- x$info$ent_ZW
    vICL <- BIC - ent_ZW
    a_tilde <- x$info$a_tilde
    return(cbind(vICL = vICL, BIC = BIC, penKG = penKG, lb = lb, entZW = ent_ZW, K = K, G = G))
  }