I want to perform a genome wide association study (GWAS) with PLINK 1.9. I have whole genome sequencing SNP calls for ~100 patients where I know in advance that there is a skew towards subpopulations of African and South American descents, with other populations represented in much smaller percentages.

I want to control for relatedness to reduce false positives, but how do I not rule out variants that are associated with the ethnicity AND contribute to the phenotype?