logLogisticRegCndtlLike: A function that evaluates the log-likelihood for a logistic...

In tbrown122387/mmcmc: Miscellaneous Markov Chain Monte Carlo Functions.

Description

This function evaluates the log-likelihood of a logistic regression model in a numerically stable way. It uses the log-sum-exp trick.

Usage

logLogisticRegCndtlLike(y, linkedMeans)

Arguments

y	observed dependent variables, assumed to be coded as 0,1.
linkedMeans	a numeric vector. The model matrix times the beta vector.

Examples

xiB <- rnorm(length(y), mean = 100)
logLogisticRegCndtlLike(y, xiB)
sum(dbinom(y, 1, inv.logit(xiB), TRUE)) #-Inf

tbrown122387/mmcmc documentation built on Dec. 31, 2019, 12:34 p.m.