We talk about unmet assumptions like assortative mating, height education etc. But as we are working with selective samples, e.g. due to volunteer bias, our samples are usually more educated, healthier, older, more urban than the general population. How does this affect our GCTA estimates?
Good question, Henrik! Indeed, ascertainment biases can affect heritability estimates and induce biases. Usually if ascertainment is directional (e.g., skewed towards people being more educated, taller, healthier or enriched with disease) then you’d expect downwards biases for traits genetically correlated with ascertainment. You may want to check out this paper for more details: https://www.pnas.org/doi/10.1073/pnas.2425530122
Note that this is not an issue with GCTA but with any heritability estimate in general. Also, biased heritability estimates are defined here relative to what you’d have from a general (unascertained) population.