Up to this point, we have described 10 separate cross-validation experiments for each HapMap 3 panel: a cross-validation within the panel of interest, followed by 9 additional experiments with successively more inclusive reference panels. We also wanted to assess the sensitivity of the inference to the khap parameter in IMPUTE2, so we repeated each of these experiments across a grid of khap values: 15, 30, 60, 90, 120, 150, 200, 250, 300, 350, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, and 2020 (which is the total number of HapMap 3 haplotypes, 2022, minus the two haplotypes of the masked individual). By definition, khap cannot exceed the number of haplotypes in the reference panel, so each reference configuration used only the values that were consistent with its panel size. Each configuration also used a special khap value that was set to the total number of reference haplotypes in that experiment, since this value seldom fell on a grid point.
