Hello. I am currently learning how to conduct GWAS analysis.
My question is about population structure in GWAS population.
For example, the phenotype in a study is stature and we want to find the underlying genes. If the population involves both Europeans and Asians, and the Europeans are actually higher than Asians.
It is common to use a linear mixed model like EMMAX. But in my analysis, the result includes many top SNPs that only belong to one population. Does it necessarily mean that linear mixed model is not perfect? How can I tell if these SNPs explain real "between group" differences (the reason why Europeans are taller) or just "spurious associations"?
I can do stratified analysis, but what if I am interested in these "between group" differences?
If most of your top hits are SNPs which are high frequency in one population and low in the other, then that definitely suggests you might not be controlling for population stratification. Always be wary of these SNPs, especially when there is a clear stratification in the phenotypes between the populations.