Confusion matrix for a logistic glm model in R. Helpful for comparing glm to randomForests.

gistfile1.r

confusion.glm <- function(data, model) {
  prediction <- ifelse(predict(model, data, type='response') > 0.5, TRUE, FALSE)
  confusion  <- table(prediction, as.logical(model$y))
  confusion  <- cbind(confusion, c(1 - confusion[1,1]/(confusion[1,1]+confusion[2,1]), 1 - confusion[2,2]/(confusion[2,2]+confusion[1,2])))
  confusion  <- as.data.frame(confusion)
  names(confusion) <- c('FALSE', 'TRUE', 'class.error')
  confusion
}

Thank you for posting...worked like a charm.

In comparing against randomForest confusion matrices, I find it easier to have true values on the left margin and predicted values on the top margin, as that's what randomForest presents. Below gives that version, with some added flexibility for providing new data or not and a customizable cutoff value.

confusion.glm <- function(model, des.mat=NULL, response=NULL, cutoff=0.5) {
    if (missing(des.mat)) {
        prediction <- predict(model, type='response') > cutoff
        confusion  <- table(as.logical(model$y), prediction)
    } else {
        if (missing(response) || class(response) != "logical") {
            stop("Must give logical vector as response when des.mat given")
        }
        prediction <- predict(model, des.mat, type='response') > cutoff
        confusion  <- table(response, prediction)
    }
    confusion <- cbind(confusion,
                       c(1 - confusion[1,1] / rowSums(confusion)[1],
                         1 - confusion[2,2] / rowSums(confusion)[2]))
    confusion <- as.data.frame(confusion)
    names(confusion) <- c('FALSE', 'TRUE', 'class.error')
    return(confusion)
}

What do you do with that, after you have run that code? There is no output, so I assume something additional coding follows this...?

In the case of a prediction vector of all True or all False [Class-Error 100%] this throws an error which is fixed by changing to
confusion <- table(factor(prediction, levels = c(FALSE, TRUE)), as.logical(model$y))

Great! Thank