I am doing a genowide association study within a self contained population. Regarding to some paper, possible population stratification can be corrected by using vectors of multiple dimensional scaling (MDS) as covariance. This can be done in PLINK. My question is, how to decide how many dimensions should I choose. PLINK give command like, --mds-plot K for such study, where k has to be define.

The other question is how to find out which vectors or how many vectors should I use as covariance? *I found a tutorial, which explains something about GenABEL, a package in R. It was written, components from MDS was plotted. There is a deep slope from the first to the second component, therefore two components are included into analysis. In my case, I have tried 10 dimensions and plot first 10 components, does not have any deep slope....

And one more question, how to integrate these vectors into PLINK linear regression analysis?

Would it be correct to integrate distances from certain dimentions for each individual as covar in PLINK?