Another advantage of our approach can be seen by comparing the running times of the restricted and full datasets for BEAGLE and IMPUTE v2. The average BEAGLE run took 3.3 times longer in the full dataset than in the restricted dataset, whereas the IMPUTE v2 running time increased by factor of just 1.1. For comparison, the total number of SNPs in the dataset increased by a factor of 3.1. This contrast between the methods arises from the way they model SNPs in U1: IMPUTE v2 models only the reference panel at such SNPs, whereas BEAGLE tries to model all individuals in the dataset. We regard the full dataset as a more realistic application of these methods, so we believe that the parenthetical running times in Table 2 offer the best comparison between BEAGLE and IMPUTE v2.
