We present a simple algorithm that uses self-reported ethnicity information, pedigree structure, and affection status to group families into genetically more homogeneous subsets. This algorithm should prove useful to researchers who wish to perform genetic analyses on more-homogeneous subsets when they suspect that ignoring heterogeneity could lead to false-positive results or loss of power. We applied our algorithm to the self-reported ethnicity information of 159 families from the Veterans Affairs Cooperative Study of schizophrenia. We compared these estimates of population membership with those obtained using the program structure in an analysis of 378 microsatellite markers. We found excellent concordance between family classifications determined using self-reported ethnicity information and our algorithm and those determined using genetic marker data and structure; 158 of the 159 families had concordant classifications. In addition, the degree of admixture estimated using our algorithm and self-reported ethnicity information correlated well with that predicted using the genotype information.