IMPUTE2 uses a Markov chain Monte Carlo (MCMC) algorithm that alternates between phasing typed SNPs and imputing untyped SNPs. Each MCMC iteration includes two steps:1.Sample a new phase configuration for each study individual, using information from other study individuals and reference panel haplotypes at SNPs typed in the study.2.Given the newly sampled haplotypes for all study individuals, treat each haplotype as independent (conditional on the reference panel) and analytically impute the alleles at untyped SNPs.This MCMC algorithm is run for a number of iterations (typically 30, including 10 burn-in iterations), then the probabilities from Step 2 are averaged across iterations to produce marginal posterior genotype probabilities at each untyped SNP.
