As both Eagle and SHAPEIT2 have parameters that trade off speed and accuracy, we also investigated the effect of running each method with non-default parameter settings. For Eagle, we benchmarked its ––fast mode, which speeds up analysis by increasing the size of SNP blocks and limiting the approximate HMM computation (Online Methods). We observed that the ––fast mode of Eagle completed analyses roughly twice as quickly as the default mode with a slightly higher switch error rate (0.321%, s.e. 0.011%) (Table 1). We also benchmarked slower parameter settings that decrease the size of SNP blocks or expand the approximate HMM computation; these modifications did not significantly improve accuracy (Supplementary Table 6). For SHAPEIT2, we increased the number of conditioning states from K=100 (the default) to K=200 or K=400, simultaneously increasing the number of ligated blocks to 4 or 5 per 10,000-SNP region to keep per-job run times within the 5-day limit (Table 1). We observed that using K=200 conditioning states achieved accuracy similar to Eagle, while using K=400 states achieved the lowest switch error rate of all methods tested (0.243%,
