Provided you believe that P(A | B) = P(A & B)/P(B), Bayes theorem always applies.

I was making fun of the previous poster, who seems to be arguing that even if group A has an average IQ of 115 and group B has an average IQ of 70, there is overlap between their distributions and so it should be shocking to you that there are more A's with Nobel Prizes and more B's unemployed or in prison.

It's not about having more or less of something. Discrimination implies an act of selection. That more A's have Nobel Prizes doesn't suggest that I should always pick A's over B's when I need an intelligent person.

Also, there are no racial groups for which your average IQs hold. My point is not just that there is any overlap, it's that there is considerable overlap, which renders discrimination based on race quite useless.

Most racists I know of don't suggest that one should use race when other variables are available -- e.g. they don't say that a black guy in a suit is more dangerous than a white meth-head with a gun, even though statistically black people are more criminal, because intuitively the armed cracked-out guy is even more criminal. I don't see how knowing average IQs of different ethnic groups obligates you to make mistaken judgments by overemphasizing them -- are you worried that other people will make this elementary mistake?

Also, discrimination means being able to tell different things apart. A 'discriminating' taste in wine means knowing the difference between good wine and bad, not thinking that all the wine from region A is better than any wine from region B.

Ashkenazim and sub-Saharan Africans?

Again, with the 'some overlap renders it useless' argument. I still don't understand why you will not also educate your child against the dangers of thinking that age correlates with height, when everyone knows that there are ten-year-olds who all taller than particular nine-year-olds -- and that you can have arbitrarily narrow age groups that make the average difference far smaller than the difference within each group. Overlapping bell curves with different medians still have different medians, however much they overlap.

My argument was in terms of real world situations in which it would be in someone's best interests to discriminate based on race. The point is that there are very few, if any, since there are other variables that are a far better proxy for intelligence than race due to the overlap in distributions. I wasn't trying to say that the overlap makes the medians irrelevant altogether; it just makes them irrelevant in almost any real situation.

The knowledge that the average ten-year old is taller than the average nine-year old is just as useless as knowing that the average Ashkenazi Jew has a higher IQ than the average sub-Saharan African.

So are you arguing that it isn't useful to know about racial differences, even if they're true? I hear about them all the time! People wonder why there are so many black people in prison, or why there are so many Asians earning PhDs. Some people have good explanations for these phenomena. One problem with ignoring data about racial differences is that you end up being forced to argue that some races (or all races) are viciously racist -- that white folks want Africa to be a hellhole, and are willing to sacrifice all kinds of natural resources to keep it that way. The world makes more sense if people vary in abilities and tendencies, rather than being absolutely equal in every way, except that some people are evil.

You would really be indifferent between finding yourself in Boro Park or Bedford-Stuyvessant? I understand the rent is very cheap in the latter (I lived there a while). Bed-Stuy doesn't have sub-Saharan Africans (or if it does, they're very smart immigrants who are likely to move to a better neighborhood in a few years), but the people there are from roughly that group, plus some admixture.

Actually, I think the reason so much attention is paid to which races earn PhDs or end up in jail is to find out how to increase the number of PhDs of other races or reduce the prison population of certain races. If those differences are caused by genetic differences, then there's nothing we can do to change that. There are other factors at work in both situations, like socioeconomic status and culture, which we can use to make everyone better off.

I think socioeconomic status is a better predictor of criminality than race. With that said, if I'm walking through an unfamiliar neighborhood, race is one of the few variables I have to work with. Unfortunately, in America, race is a moderately effective indicator of socioeconomic status, so I would use it as one factor in judging my safety. Even then, I usually use clothes and appearance rather than race to judge one's propensity for crime.

This is actually a more interesting problem in terms of what to tell children than racial differences in intelligence. I think by the time they're old enough to be somewhere without adult supervision, they are able to determine what looks like a poor area or poor person and act accordingly. I can see the logic that would lead someone to tell their kids to use race as such a proxy instead, but the consequences of doing so make it not worth the negligible increase in safety in my opinion.

Actually, race is the single best predictor of how much crime to expect in a given area. Besides population already in prison, at least.

I think it's a red herring to ask why we're collecting the data in order to decide whether or not the data are accurate.

Really? That would be interesting data to look at. Do you have a source?

I don't care why we're collecting the data, and I was assuming the data was accurate. I'm just saying that as far as genetic differences are a factor in the data, there's nothing we can do about it or with it, so it's useless.