The flow of the haplotypic analysis was as follows: CHM was used in a sliding window fashion, using window sizes from one to 12 SNPs (W = 1, … ,12), with the maximum size pre-chosen to be equal to one-half of the number of markers. Window size of one corresponds to the test of an additive effect of an SNP (‘allelic trend test’). The permutation-based overall test (100 000 permutations) of composite allelic association was computed for each window, the –ln(P-value) plotted and compared with the conservative Bonferroni threshold that accommodates the number of tests in each window. The tests involving the different W are not independent, because they contain and share linked markers in LD. Nevertheless, if significance of association persists across windows of different sizes (as we indeed observed), the overall confidence in the association should be increased.
