Performs likelihood ratio test on `logLik` objects.

lrt.R

#' @param \dots two or more `logLik` objects
lrt <-
function(...) {
    logL <- list(...)
    nmodels <- length(logL)
    rval <- matrix(rep(NA, 5 * nmodels), ncol = 5)
    colnames(rval) <- c("#Df", "LogLik", "Df", "Chisq", "Pr(>Chisq)")
    rownames(rval) <- 1:nmodels
    rval[, 1] <- as.numeric(sapply(logL, function(x) attr(x, "df")))
    rval[, 2] <- sapply(logL, as.numeric)
    rval[2:nmodels, 3] <- rval[2:nmodels, 1] - rval[1:(nmodels - 1), 1]
    rval[2:nmodels, 4] <- 2 * abs(rval[2:nmodels, 2] - rval[1:(nmodels - 1), 2])
    rval[, 5] <- pchisq(rval[, 4], round(abs(rval[, 3])), lower.tail = FALSE)
    rval
}