Is it "statistically legal" to perform a t-test on the averages of two groups? In other words, the data set values are averages themselves (example: a ttest performed on GPAs). I realize that doing a ttest on averages is pretty much like taking the average of averages, which is in most cases a...