Predicting IQ from genetic information alone should remove all the of compounding factors (except those introduced by IQ measurement itself) but can only serve as a lower bound unless predictor overfitted the study data.
> The truth is most likely somewhere between those two.
> but can only serve as a lower bound unless predictor overfitted the study data.
Yes, but the truth is likely somewhere much closer to the upper bound (twin studies) than the lower bound (genome studies).Genome studies are only good for lower bounds. They're a new methodology that isn't good with complex traits. For example, they can only explain 20-30% of the variability in hair color even though we know that hair color is almost 100% genetic due to the observation that identical twins always have the same hair color. Intelligence is much more complex than hair color, so it's no surprise that this method has failed to explain more than 5%. Further, these studies often rely on weak proxies like educational attainment because of the difficulty of IQ testing at scale, and the studies that do try to IQ test at scale suffer from small sample sizes which impacts the ability to find results due to being underpowered.
Also, IQ becomes more correlated with parents as you age, suggesting childhood environment plays a less important role.
In addition, the jump in correlation from fraternal to identical suggests more role for genetics than merely a few %.
> any claim about the genetic component kinda falls apart for me based on them as the children were in the same pregnant mother at the same time.
They do somewhat control for pregnancy by comparing identical vs fraternal twins, and finding a higher correlation in intelligence between the former, as well as a far more similar brain structure in regions known to be related to intelligence (frontal cortex).
Anyway, it's impossible to get rid of all confounds via twin studies. But this doesn't mean that twin studies provide zero evidence. It just means the evidence can't be fully conclusive. It just strongly points in the direction of genetics.