I can’t say much beyond that now because I’m swamped, and can’t spend the time finding the original (must be a PDF around here somewhere!) It is not open access, they must have sent it to me.

My impression at the time was not that this was an invalid statistical approach.

Again, it is not a 500,000 dollar study of a drug trial. It is a study of a very limited population (how many adult film actresses are there?) and strong differences were found with the background non-adult-film-acting-women population, which was a reasonably sampled with an appropriate methodology (matched pairs).

Not sure what you mean by matched paired sampling here. That process typically involves randomly assigning one member of each pair to a treatment group (as in sure you know, so don’t take that as a lecture on basics). Maybe you’re intending case control or some use of matched pairs I’m not familiar with.

Still, without a proper bit of random selection and or assignment, you can still do the calculations but the stuff needed to justify rhe work leading to inference isn’t met.

Research in psychology and sometimes sociology (and anthropology) is simply not conducted the same way as, say, drug trial research. Only drug trial research is conducted like drug trial research, yet other forms of research are often held up to the same standard. This is an example of that.