Although the exact algorithm is unpublished, according to the documentation, PLINK appears to directly estimate haplotype frequencies using an expectation-maximization algorithm based on a multinomial model [Purcell et al., 2007]. We used default parameter settings as follows: selecting at most 5 proxy SNPs, searching up to 15 SNPs around the index SNP, searching within 250 kb around the index SNP, genotype missingness of ≤0.2 at proxy SNPs, and a minor allele frequency of ≥0.005 at proxy SNPs. PLINK produces an output file containing the posterior probabilities for the three genotypes at each SNP for each individual. Software is available at http://pngu.mgh.harvard.edu/purcell/plink/.