Consider finding 3: "Heritability is caused by many genes of small effect"
This is exactly what you would expect to find if you do a regression with 1000s of variables: many effects, mostly small ones.
Of the 10 findings only 1 is clearly not a fact about statistics in settings where dimensionality is very high relative to sample size.
Finding 5. The heritability of intelligence increases throughout development
Observe that this finding is based not on associations between particular genes and traits, but instead on twin studies. Twin studies have their own problems when it comes to causal identification.
That said, the overall claims here make sense and have some empirical support. My point is that the field has long made outsized claims based on dubious science that is legitimized by it's appropriation of genetics (even though it is a branch of psychology, not biology).
Panofsky's book "Misbehaving Science: Controversy and the Development of Behavior Genetics" is excellent.
A replication is when one person/group does something, they write down instructions for someone else to do the same thing, and then the results are compared. Is there a single mention of this happening in that paper? I expected a table of such results.
Ie, replication is about doing some very specific thing, instead I see a bunch of vague stuff like "abnormal is normal"...
No, GWAS is smarter than that.
The article mentions multivariate regression quite frequently—does anyone know if more recent machine learning techniques have made their way into the field?
Oh god I hope not.
> Finding 2. No traits are 100% heritable
Although heritability estimates are significantly greater than 0%, they are also significantly less than 100%. As noted earlier, heritability estimates are substantial, typically between 30% and 50%, but this range of estimates is a long way from 100%.
Besides, some of the findings look significantly over generalized, and the ones that aren't don't seem very revolutionary or intetesting.
While there is obviously a non-shared environmental component, it is really noise. The only factors that people care about are the factors that can be studied; shared environment (that is all the environmental factors that can in theory be measured) and genes. Non-shared environment is just what is left over once you account for measurable environmental factors and genes.
I've browsed the paper, but must of missed it. How big are the error bars? Is it possible to get a sense of how much variation is explained by genetics alone?
The error bars depend on the trait being measured. Some are relatively small (like g), while others like "agreeableness" (just choosing a random trait) will be quite large as it is hard to measure this accurately and consistently.