I want to take a dataset that has truth and various predictors in it and summarise the auc, ‘best’ threshold, sensitivity, and specificity for each predictor, split by some grouping variable.