Finally, we investigated the utility of an African American reference panel, the HapMap phase III ASW data. The HapMap phase III ASW, phase III CEU, and phase III YRI reference panels contain different numbers of SNPs. To enable a fair comparison among reference panels controlling for SNP density, we first extracted a common set of 12,941 reference SNPs shared across all three panels. Using this subset of shared SNPs, we determined imputation yield using the fixed error rate approach. The phase III YRI reference panel slightly outperformed the phase III ASW reference panel (Table V). We also estimated the number of SNPs in approximate linkage equilibrium by pruning based on pairwise correlation. This analysis revealed that the number of SNPs in approximate linkage equilibrium was much higher in the phase III YRI reference panel than in the phase III ASW reference panel, but that controlling for an equal number of founders in both reference panels eliminated most of this difference (Table VI). The results imply that, on average, the levels of linkage disequilibrium (and consequently, the average lengths of haplotype
