One drawback of using Tukey's HSD is that adding multiple predictors and interaction terms.... O creates models that do not converge O is that type II error increases O is there are not enough comparisons O creates an unwieldy amount of comparisons