Using the same approach, we also checked the dependence of imputation accuracy on pedigree information, as trios improve the accuracy of haplotype phasing and therefore imputation. We compared the within-CEU results described earlier to imputation done purely within the TSI sample, with the sample size held fixed. The two populations are closely related, but the CEU samples were genotyped in trios and the TSI samples as individuals. The results were virtually identical (data not shown), indicating that poor phasing was not a problem for our unrelated samples, at least for array SNPs. (Note that pedigree information was used indirectly in our TSI phasing, with phased CEU chromosomes used as a reference panel for phasing TSI.)
