This is an Error, Right?

[Jman28]

I just started reading 'True Odds' by James Walsh. It seems like it's gonna be a good read. It is about identifying mistakes people make in gathering, interpreting, and analyzing statistics to determine risk in their life.

About 30 pages in I come to what I think is an error. Normally I would just move on I guess, but in a book specifically about correcting statistical errors, a minor mistake in exactly that hurts the books credibility, in my opinon.

Anyways, here's the mistake. Let me know if I'm wrong.

He has the stat that in a given year, the odds that a gay man will contract AIDS is 1 in 245. Since condoms are 90% effective in blocking the transmission of AIDS, he reasons, using a condom every time you have sex will improve a gay mans odds to 1 in 2450.

My main problem with this is that it assumes that no gay man ever uses a condom, clearly a bad assumption.

Also, doesn't the fact that condoms are 90% effective each time used not translate into 90% year long reduction?

Should I even care if this is a mistake?

[Good Friar]

The assumption does seem bad. The 90% statistics cited with condoms are usually added with the caveat of something to the effect of "within a year of continuous use", and doesn't really show the effectiveness in a given use. To get a legitimate experimental probability, they'd have to have people knowingly have sex with an infected partner, which is obviously unethical. Medical science and physiology often has very poor data because of such complications.