R/getNegativeLogLikelihoods.R

In HCBravoLab/metagenomeSeq: Statistical analysis for sparse high-throughput sequencing

#' Calculate the negative log-likelihoods for the various features given the
#' residuals.
#' 
#' Maximum-likelihood estimates are approximated using the EM algorithm where
#' we treat mixture membership $delta_ij$ = 1 if $y_ij$ is generated from the
#' zero point mass as latent indicator variables. The log-likelihood in this
#' extended model is $(1-delta_ij) log f_count(y;mu_i,sigma_i^2 )+delta_ij log
#' pi_j(s_j)+(1-delta_ij)log (1-pi_j (sj))$. The responsibilities are defined
#' as $z_ij = pr(delta_ij=1 | data and current values)$.
#' 
#' 
#' @param z Matrix (m x n) of estimate responsibilities (probabilities that a
#' count comes from a spike distribution at 0).
#' @param countResiduals Residuals from the count model.
#' @param zeroResiduals Residuals from the zero model.
#' @return Vector of size M of the negative log-likelihoods for the various
#' features.
#' @seealso \code{\link{fitZig}}
getNegativeLogLikelihoods <-
function(z, countResiduals, zeroResiduals){
	pi=getPi(zeroResiduals)
	countDensity=getCountDensity(countResiduals, log=TRUE)
	res=(1-z) * countDensity
	res=res+sweep(z, 2, log(pi), FUN="*")
	res=res+sweep(1-z,2,log(1-pi), FUN="*")
	-rowSums(res)
}