Haplotype blocks were created if 95% of informative marker comparisons were in “strong LD,” which is defined by the pairwise 95% confidence bounds by the algorithm described in (Gabriel et al., 2002). Haplotypes were tested for association using adjusted haplotype regression models in PLINK, which assigns a probabilistic set of haplotypes for each individual. (Purcell et al., 2007). PLINK was also used to test for population stratification by grouping the sample according to genome-wide average identity by state sharing at more than 32,000 markers spaced across the genome. Stratification is assumed to be present if multiple groups are identified.
