Tuesday, November 30, 2010

A new study that says gay people are coming out earlier than in the past . . .

. . . is wrong in a "new and interesting" way.

That blog post (1) explains how the study went wrong and (2) asks whether it could have possibly been right.

As to the second point, the blogger (Ben Goldacre) explains:

It’s a difficult analysis to design, because in each age band, there is no information on gay people who are not yet out, but may come out later, and also it’s hard to compare each age band with the others.
(The comments section on that post also has a lot of relevant insights.)

This reminds me of the oft-repeated factoid that "50% of marriages end in divorce." How could you ever determine whether this is true? You can observe divorces that have actually happened, but you can't possibly know whether existing marriages will end in divorce.

Even questions that seem to be about concrete, observable facts can't necessarily be answered by empirical research.

4 comments:

Thorne Cassidy said...

I have to disagree with your logic:

"the oft-repeated factoid that '50% of marriages end in divorce.' How could you ever determine whether this is true?"

This is an analysis of past data to arrive at some useful tool for the present in the same way that life expectancy is. It's reasonable to come to such conclusions. Now, as a philosophy major, I'm sympathetic to an argument against all inductive reasoning: "just b/c the sun arose yesterday, doesn't mean that it will tomorrow..." But then you'd have to dismiss the fundamental axioms of philosophy and of reason themselves b/c even though they were true in the past... I'm a pragmatist when it comes to actually living my life--AND SO ARE YOU and virtually everyone else. It's fair to gamble that in the current cultural milieu, with all things being equal (when are they?), that there's a 50% or better chance of divorce. Couples will have to work on the other controllable factors to avoid it.

John Althouse Cohen said...

I agree with your abstract points about inductive reasoning. I just don't think the statement that "50% of marriages end in divorce" is a good example of inductive reasoning. Actually, it seems that you agree with me on this, since you only say you'd "gamble" that at least 50% of marriages end in divorce. That's just a guess, not a legitimate statistic.

Thorne Cassidy said...

Sure, I guess. We did use the term "statistic." But your objection was that you can't conclude the longevity of existing marriages from observed divorces that have happened. right? That's a denial of induction. I'm assuming your referring to legally recognized--and documented--marriages and divorces. These are verifiable statistics used for prognosticating--always a best guess as most science is. Let's call this one a draw. LOL Take care.

John Althouse Cohen said...

But your objection was that you can't conclude the longevity of existing marriages from observed divorces that have happened. right? That's a denial of induction.

As you recognize, the only way divorce statistics could be meaningful is if they're based on marriages from the past. You can't look at the people who have gotten married this year; they haven't had any time to get divorced. By the same token, looking at marriages from the past 5 years isn't adequate either, since plenty of those people might still get divorced in the future. Same thing if you go back just 10 years. How far back would you go? You'd need to go back at least a few decades (say, 40 or 50 years) to feel confident that you were observing just about all the divorces — not only the ones that have happened so far, but all the ones that will happen. But if you go back that far, why would you assume that the divorce rate has anything to do with the present likelihood of a married couple getting divorced? The conventional wisdom is that divorce is skyrocketing. You can't just dismiss that possibility by saying divorce is like the sun rising.

So, I'm denying whatever induction was used to arrive at this statistic because I don't think it follows. I'm not trying to make a philosophical point about induction. But if you think induction in general is no better than the induction that supports (or, fails to support) the divorce factoid, then so much the worse for induction!