Sparse principal component analysis (sPCA) was used to determine genotypic profiles which discriminated between modalities and sexes. In contrast to the tests for significant SNPs, sPCA identifies only individual SNPs which have differential effects between sexes, modalities, or localities. Parameters for sparseness were two components with no more than 10 SNPs per component applied to the aggregated effect sizes described in section 8.3.2. Given the restriction of 10 SNPs in each component, two-thirds of the 147 SNP-age range combinations never appear in the first component. This is the non-significant group for determining the empirical null distribution. The significance of the members of the group non-zero components are evaluated based on the empirical null distribution. The same sPCA analysis was applied to the 1000 bootstrap samples described in section 8.2 and the proportion of occurrence of each of the 147 SNP-age range combinations in the first sPCA component of the 1000 bootstrap samples determined, as well as the frequency of occurrence of the significant and non-significant SNP-age range combinations from the actual sample. The 95th percentile of cumulative distribution function of
