Can I conclude whether these two populations are different?

Question

I'm trying to determine whether my patient population differ from my control population in regards to number of children fathered per individual.

The tables show the frequency of amount of children fathered after operation for the patient population and control population respectively.

First I calculated confidence intervals the old fashioned way using SAS, but since my data is not normally distributed I came to the conclusion that this was not the way to go. I also tried bootstrapping the mean and confidence intervals and found the exact same confidence intervals as with the standard method.

Am I going about this all wrong?

Answer 1

First, I don't think you need any test here. With samples of this size, even a tiny difference is likely to be significant, but it's not going to be meaningful. Look at the %ages in each table! They are virtually identical!

Second, you didn't say what you did to compare the means. You just said "the old fashioned way". I don't know what that means. You found confidence intervals but it seems like you want some sort of statistical test.

Third, if you insist on a test, I wouldn't use t-test (see Should I use t-test on highly skewed and discrete data?) and certainly not chi-square. I'd use something that accounts for the fact that the variable is a count (that is, a nonnegative integer). One route is to use Poisson or NB regression. Since you are clearly a SAS user, I'll note that these are available in PROC COUNTREG.