How do you define racial groups? its something that constantly shifts. I have an encyclopedia for the 1950s that defines South Asians as "Caucasoid" which is not how most people now think.
https://upload.wikimedia.org/wikipedia/commons/f/f9/Principa...
Obviously some racial groups which are heavily socially constructed are not borne out by the data. There's no cluster for 'Latinos'. But in general I think there is quite a high correspondence with race as it is constructed in society and culture.
There are certainly some cases where the genetic data matches conventional thinking about race, but also many where they flatly contradict. It also suggests that saying something like "white people are genetically smarter on average than black people" is highly suspect. The group of people that we call "black" has a greater range of genetic profiles than the rest of the world put together, and the group we call "white" is kind of confused—to be genetically meaningful it would probably need to either include North Africans and Arabs or exclude Southern Europeans.
Plus, none of these studies actually spend the money to do any genetic testing, they simply assign people to "white" or "black" race based almost always on skin color (which, since the gene for dark skin is dominant, is a very poor marker of actual genetic similarity - the great grandchild of a dark-skinned person with otherwise only white-skinned relatives can still be dark-skinned).
Edit: and I'm not even mentioning the obvious publication bias, as essentially only racists care about this race vs intelligence discussion, at least enough to invest in creating studies about it. Plus the fact that IQ itself is actually a very problematic measure, when applied beyond its actual medical use (quantifying mental disability).
What if some of the observed racial differences were caused by a lack of vitimin D, due to darker skin photosynthesising less of it? Then we could fix the prison crisis with a pill!
If we block all research into a problem then we can’t solve it.
If you actually care about fixing a problem then we should bring all available tools to fix and solve it.
Correlation measures whether Y=mX+c holds with small error. I sometimes think that many people unknowingly assume that all dependencies are somehow straight-line ones. This is distinct however from "linear approximations" in physics, which can at least be justified sometimes.
I also suspect that correlations of 0.5 or lower are often just statistical noise.