To check for association between case-control status and population structure, we test for an association between disease status and the first two eigenvectors from the PCA of each population subgroup of interest. No significant associations have been found so far, which may be a reflection of study designs that carefully match the geographic origins and other characteristics of cases and controls. We also test for a difference in missing call rate per sample between cases and controls, as a way of detecting association with experimental factors. No significant differences have been found so far, except for a special case in the Addiction project described below. In addition to case-control status, we test for correlation between missing call rate and quantitative traits of interest. For example, in the T2D NHS project, the correlation between body mass index and missing call rate is small and not significant (r=−0.01, p-value=0.28).
