Type I Error Rates for Unrelated Samples and Nuclear Families.

Notes: In homogeneous and admixture population settings, we sampled 400 unrelated subjects and 200 nuclear families when the binary trait was not involved, and sampled 200 unrelated cases, 200 controls, and 200 nuclear families with at least one affected child when the binary trait was involved. In stratified population settings, we sampled 250 unrelated subjects and 150 nuclear families from population A, and 150 unrelated subjects and 50 nuclear families from population B when the binary trait was not involved. When the binary trait was involved, we sampled 150 cases, 100 controls and 150 nuclear families with at least one affected child from population A, and 50 cases, 100 controls and 50 nuclear families from population B. Type I error rates for univariate and bivariate analyses are estimated for the combined data of unrelated samples and nuclear families under homogeneous, stratified, and admixed populations. The error rates are estimated on 1,000 replicates.a: ��-��, for EIGENSTRAT, only univariate analyses are available.Abbreviations: T, the proposed test applied to the combined sample; TU, the proposed test applied to unrelated sample only; TF, the proposed test applied to nuclear families only; ESTRAT, the method proposed by Price et al. [6] and implemented in the software EIGENSTRAT, applied to unrelated sample only; FBAT, the program FBAT [48]; Fisher, the fisher product test on the outputs from EIGENSTRAT and FBAT.