Consider a population-based sample of cases and controls, which we expect to be ascertained from a genetically homogeneous population, or genetically frequency matched for ethnicity or location in a potentially heterogeneous or admixed population. We assume that we have a series of genetically unmatched external samples that are used to expand the control group in our goal to increase power. All samples are genotyped for the same N SNPs, genome-wide. We develop a statistical method to test for association of genetic markers with disease by comparing genotype frequencies between cases and the expanded control cohort that corrects for population structure, when it exists, thus retaining the correct false-positive error rate.
