logLikNB: Log likelihood for Negative Binomial

In azhu513/apeglm: Approximate posterior estimation for GLM coefficients

Log likelihood for Negative Binomial

Description

This is a simple function to be passed to apeglm as a log likelihood for the negative binomial distribution.

Usage

logLikNB(y, x, beta, param, offset)

Arguments

y	the counts
x	a design matrix
beta	the coefficient vector (natural log scale)
param	the overdispersion (alpha in DESeq2 notation)
offset	an offset matrix (natural log scale)

Value

the log likelihood for each sample in y

Examples

# this function is used by 'apeglm' to specify
# a negative binomial log likelihood.
# so it's only passed as an argument, not for use on its own.
# we can show its output nevertheless:

y <- rnbinom(10, mu=100, size=1/.1)
x <- cbind(rep(1,10),rep(0:1,each=5))
beta <- c(log(100),0)
param <- .1
offset <- rep(0, 10)
logLikNB(y, x, beta, param, offset)

azhu513/apeglm documentation built on June 13, 2024, 3:37 a.m.