Sometimes on TRP, TBP or here, I see statistics like, "If you marry someone of age such-and-such" or "if you marry someone with such-and-such number of sexual partners" then you'll have "more chance of being divorced" or some such event. The numbers and the events differ, but the treatment of the statistics always seems to suggest a change in probability.

But isn't that wrong? Let's say that the findings are that people who marry women who are under 20 then you are less likely to get divorced (I don't know if this is a real statistic, mind). This doesn't mean that by marrying someone under twenty you've decreased the probability that you'll get divorced. The findings mean that of so many surveyed people, there were certain correlations between them.

It could be, for example, that someone who is likely to marry a younger women is possibly less predisposed to be a certain type of character, and that those characteristics imply a lesser chance for divorce. If something like that is the case, then someone who wasn't predisposed to marry a younger women wouldn't necessarily become a different character as a result, and might not change their chances of divorce at all. Acting on these statistics as if they were probabilities doesn't change anything, in that case.

This doesn't just apply to divorce, but also marriage, sex, LTRs, whatever - I constantly see statistics being interpreted as probabilities.

If that's the case, shouldn't we take statistics with a few more grains of salt than we currently take them with?