Imputation using BEAGLE [24] was conducted using all default options. When using MaCH [5], model parameters (crossover and error rates) were estimated prior to imputation by randomly selecting a subset of 200 haplotypes from the study subjects and running 100 iterations with the command options –compact and –greedy. Genotype imputation was then carried out using the model parameter estimates from the previous round with command options of –compact, –greedy, –mle, and –mledetails specified. When using MaCH-Admix, the step 1 imputation settings included: –runmode EstimateParameterOnly -r 30, –fittingstates 200, and –autoflip. MaCH-Admix step 2 setting included: –runMode ImputeOnly, –compact, and –autoflip.
