Given the need for computational efficiency with the large number of imputations (10 repetitions for each imputation procedure), our analyses focused on chromosome 22, as other studies evaluating imputation performance have done [15], [29], [32]. To ensure comparability of imputation results across chromosomes, we conducted imputation using MaCH on chromosomes 1 and 22 with reference to YRI from HapMap phase III. Imputation performance was slightly better on chromosome 1 (concordance = 92.1% and average r2 = 0.83) than on chromosome 22 (concordance = 89.2% and average r2 = 0.80), likely due to a larger number of SNPs available for comparison and slightly higher linkage disequilibrium levels on the larger chromosome.
