R/loglik.R

In RAMP: Regularized Generalized Linear Models with Interaction Effects

loglik <- function(X, y, beta, family) {
    link = as.vector(X %*% beta)
    n = length(y) 
    if (family == "gaussian") 
      return(n * log(mean((y - link)^2)))
    if (family == "poisson") 
        return(-2 * sum(exp(link) + 2 * y * link))
    if (family == "binomial") 
        return(2 * sum(log(1 + exp(link)) - y * link))
       
}