Dymanic said:
How else could we confirm a thing like that, if not on a statistical corelation of some type?
Probably not a useful answer to your question, but two things to look out for in statistical studies:
1. Were the subjects assigned randomly to the different groups? You can easily do this, for example, when testing for the efficacy of drugs. One group gets the drug, the other gets a placebo. If you do this, and get a statistically significant result, you may reasonably state cause and effect, e.g. taking the drug makes you better. It can be argued that bad luck meant that healthier people were put in the drug-taking group, but that can be accounted for by tests of significance.
In a study of something like homosexuality, however, you can't assign people to the homosexual and non-homosexual groups. Even if you had some special ray-gun which turned people gay, you would probably run foul of ethics committees before getting your hands on the cash to carry out the experiment. Here, you can only do a retrospective study, and you can never be sure that there isn't some hidden variable at work that you've not identified. At best you can only say the correlation is suggestive, but you can't say what's a cause and what's an effect (maybe fingertip ridges can turn you gay?) - you may have two effects.
Of course, problems with causation vs correlation can also apply to randomised studies. If you have a plausible mechanism to take you from cause to effect, you're on much safer ground.
2. Were the test subjects selected randomly from the general population, or from some subset? If they weren't from the general population (which, let's face it, is extremely difficult to arrange), then strictly speaking,it's only valid to apply your conclusions to the population from which you sampled. However, in many cases, the results would be extremely suggestive in applying to the population as a whole.