Figure 1 also shows how imputation accuracy improved as particular panels were added to the reference set. The trends for the ASW panel (Figure 1A) are basically as expected: each successive African panel (MKK, LWK, and YRI) improved the accuracy incrementally, and the panels that capture components of European ancestry (MXL, GIH, TSI, and CEU) collectively raised the accuracy to its maximum value for this experiment. The east Asian panels (CHB+JPT and CHD) did not increase imputation accuracy any further, but nor did they reduce it; this demonstrates the ability of the method to ignore unhelpful reference haplotypes.
