The power of genetic diversity

[Pat]

That would seem to be the major divide, yes. But of course everything comes down to how you define race.

[grumbler]

Grumbler what do you have to say about 1-4% of eurasian genome coming from neanderthals? Isn't 1-4% a big enough difference for different races to exist?

I say 'that's interesting" but don't see its relevance for the discussion of "race."

[grumbler]

That would seem to be the major divide, yes. But of course everything comes down to how you define race.

Or whether you define it.

[garbon]

The misuse of "race" in medical diagnosis is a known problem (http://www.pbs.org/race/000_About/002_04-background-01-y.htm is from a doctor talking about exactly that).  The problem with the assumption that, because some doctors sometimes do use "race" in diagnosis, "race" must have medical utility is that it does not.  It is assumed to have utility, and doctors, like posters on internet discussion boards, challenge their own assumptions all too rarely.

Yeah actually the OP of this thread seems to admit that. After all, if "race" was a useful proxy, just find "one person with one Korean parent and one white parent" and she'd be good to go.

[DGuller]

Yeah actually the OP of this thread seems to admit that. After all, if "race" was a useful proxy, just find "one person with one Korean parent and one white parent" and she'd be good to go.

  That's not the necessary implication at all, that's a gross misunderstanding of proxy variables.

Hypothetically speaking, let's say that you need a new kidney.  For whatever strange reason, a kidney from a blond guy has a 1% chance of being a good match for you.  A kidney from a non-blond guy has a 0.01% chance of being a good match for you.  We don't know why hair color makes such a difference, there seems to be no causative reason for that, but no matter what other reasonably observable variables we use, we can't find a set that would explain the same piece of variance in results that hair color does. 

Therefore, hair color is a good proxy variable for screening potential match donors.  If you were looking to search for strangers that could be a good match for you, obviously it would be a good idea to concentrate your limited resources on just blond strangers, since a blond stranger is 100 times more likely to be a match.  That's an astonishing predictive power, and that makes hair color an extremely good and useful proxy variable.  However, it doesn't mean that the first blond guy you come across would be a good match for you.

[The Minsky Moment]

"Caucasian" is definitely my favorite racial categorization.  Nothing like using the name of a Central Asian mountain range to categorize a bunch of otherwise totally unrelated human beings.

[LaCroix]

Grumbler what do you have to say about 1-4% of eurasian genome coming from neanderthals? Isn't 1-4% a big enough difference for different races to exist?

now what sort of race would that be, of schizophrenia? 

but no, you're playing with science and using it to try and fit in this notion of race you seem obsessed with

Used to be a controversial minority view though. Everything is evident in hindsight, but yeah, it did make sense all along, I guess.

i know. a friend who dabbles in genetics tried telling me otherwise years ago, and then deflected my question on gingerism. i had the last laugh there

[grumbler]

Hypothetically speaking, let's say that you need a new kidney.  For whatever strange reason, a kidney from a blond guy has a 1% chance of being a good match for you.  A kidney from a non-blond guy has a 0.01% chance of being a good match for you.  We don't know why hair color makes such a difference, there seems to be no causative reason for that, but no matter what other reasonably observable variables we use, we can't find a set that would explain the same piece of variance in results that hair color does. 

Therefore, hair color is a good proxy variable for screening potential match donors.  If you were looking to search for strangers that could be a good match for you, obviously it would be a good idea to concentrate your limited resources on just blond strangers, since a blond stranger is 100 times more likely to be a match.  That's an astonishing predictive power, and that makes hair color an extremely good and useful proxy variable.  However, it doesn't mean that the first blond guy you come across would be a good match for you.

You haven't made any argument (nor has anyone else) that any of this has to do with "race," BTW.  Blonde hair has nothing to do with "race," insofar as I know.  That genetics has to do with suitability as a donor is rather tautological, I would think.

I would also argue that, since hair color can be changed, it isn't a particularly good proxy variable for screening potential match donors.  Having "blonde" genes might be a good proxy except that if one is checking genes one might as well just check for the genes that are directly applicable.

[DGuller]

You haven't made any argument (nor has anyone else) that any of this has to do with "race," BTW.  Blonde hair has nothing to do with "race," insofar as I know.  That genetics has to do with suitability as a donor is rather tautological, I would think.

I was making a point about proxy variables in general, and deliberately used a hypothetical example that had nothing to do with race.

I would also argue that, since hair color can be changed, it isn't a particularly good proxy variable for screening potential match donors.

You seem to be unduly fixated on the need for proxy variables to be very precisely defined, and not being controllable.  That's preferable, but far from necessary.  In my line of work, plenty of variables are proxy variables that are controllable by the insured, or not defined in an ironclad way.  They're still very useful, and ignoring them would lose you big money when you try to write insurance.

Here is a very basic concept:  if a variable is not a good proxy variable, then you would not observe correlations involving that proxy variable.  If you observe a correlation with some variable, however loosely or unreliably defined, then by definition you have a useful proxy variable.  If the proxy variable was useless, you wouldn't find a correlation in the first place.  It doesn't mean that you can't find an even better correlation if you sit down, and try to come up with a better definition, but what you have now is already useful.

[The Minsky Moment]

The problem DG is that traditional racial categories usually don't serve a very good proxies for the kind of genetic matching you are referring to.  That is because the construction of racial categories preceded the science of genetics and was done for very different purposes.

[The Brain]

DG, it's no use. It's been 8 years since I discussed race with grumbler and realized that he is a member of the Moron race. He won't understand, he is too stupid.

[grumbler]

I was making a point about proxy variables in general, and deliberately used a hypothetical example that had nothing to do with race.

Okay.  I would argue that your statements about proxy variables are so vague as to be trite, then.

You seem to be unduly fixated on the need for proxy variables to be very precisely defined, and not being controllable. 

You seem to be unduly fixated on avoiding any definition of "race" whatever.  I am not concerned with, let alone fixed on, a need for "race" to be "precisely defined;" just that it be defined at all. So far, you arguers in favor of the concept of "races" have signally avoided any effort to define your terms.

[grumbler]

DG, it's no use. It's been 8 years since I discussed race with grumbler and realized that he is a member of the Moron race. He won't understand, he is too stupid.

Thanks for another juvenile attempt to derail an intellectual discussion with an ad hom.  DG and I are too adult, alas for you, to fall for such a feeble troll. 

[The Brain]

DG, it's no use. It's been 8 years since I discussed race with grumbler and realized that he is a member of the Moron race. He won't understand, he is too stupid.

Thanks for another juvenile attempt to derail an intellectual discussion with an ad hom.  DG and I are too adult, alas for you, to fall for such a feeble troll. 

[DGuller]

The problem DG is that traditional racial categories usually don't serve a very good proxies for the kind of genetic matching you are referring to.  That is because the construction of racial categories preceded the science of genetics and was done for very different purposes.

That's a question that can be answered by statistics.  I do know that some statistics (namely life tables) produce a shockingly different results for whites and blacks.  However race was defined when life tables were produced, it was defined well enough that you're getting some really enormous differences.

For example, at age 5, life expectancy for a white person was 73.4 in 2005.  For a black person, it was 69.0.  (I'm using life expectancy at age 5 to take out the effect of higher child mortality rate for blacks, though that by no means is an issue that should be ignored either).  http://www.cdc.gov/nchs/data/statab/lewk3_2005.pdf

Now, you may ask what in this case race is a proxy for, and that would be a very good question.  It could be a proxy for socioeconomic class, it could be a proxy for biological differences, something else entirely, or a combination of the above.  However, you cannot say that race is not a good proxy variable in this case, 4.4 years in life expectancy is a big difference that is explained by this allegedly poor proxy.