Should Covariates be log transformed in a GWAS study

Question

[mattmacknev]

Hi, I am analysing some quantitative traits on an Exome Chip SNP dataset in PLINK and RareMetalWorker. I am Ln transforming my Quantitative traits but I am not sure whether I should do the same for my covariates. For example, If I am looking at regional fat distribution, age and % total-fatmass should be added as Covariates to adjust for overall fatness and increased fat with age. The regional fat measure is Ln transformed, so should I therefore also Ln transform my %Total-fatmass covariate? What about Age?

[jiangjing_ing]

Hi, I have the same question. Did you find out that? Could you explain the problem to me? Thx!

Answer 1

[Kevin Blighe]

As I mention here: A: Log-tranformation and GWAS

You don't have to log everything. You just have to ensure that each variable has a distribution that is suitable to the assumptions of the statistical test that you are aiming to employ. Usually that means that they should be normally distributed. You do not have to log everything, though.

Kevin