A standard likelihood ratio test was used to determine whether haplotypes constructed from the haplotype blocks were better able to explain the trait variation then the most associated SNP in the block. First, we regressed the trait value on predictors that count the number of each haplotype possessed by a participant. Individuals with rare haplotypes had the count for that haplotype equally distributed among the other haplotypes. This is akin to the method of Zaykin et al. (Zaykin, Westfall, Young, Karnoub, Wagner, & Ehm, 2002). Next, we regressed the trait value on the allele count of the most significantly associated SNP in the haplotype block. Multiplying two times the log ratio of the haplotype regression likelihood and the SNP regression likelihood produced a statistic that is asymptotically chi-square distributed with degrees of freedom equal to the number of unique haplotypes minus the number of unique genotypes at the SNP.
