Hi all,

Recently, I read quite a few papers in the GWAS field and noticed something which I did not quite understand. When looking for associations, often a "simple" linear model is used, however, before fitting the model, many authors correct the "outcome" variable for covariates such as age, batch, ... and then use the corrected values (or residuals) as the new "outcome" measure. Here ist just one example of such a paper (see "Proteomic profiling"). Previously, I have done a lot of differential gene expression analysis, and there, the "best practice" is to simple include your covariates in your GLM formula (outcome ~ SNP + age + batch + ...) .

I don't quite understand why people in the GWAS field are not doing the same, given that they also use linear models.

Any insights or pointers to papers are much appreciated!

Cheers!