To characterize the age specific variation of genotypic effects on the phenotypic variables, significant SNPs are identified on an SNP specific basis. Two complementary methods are used, estimation of the false discovery rate using the method outlined in Storey and Tibshirani (2003), and the assessment of the concentration of effect sizes in particular SNPs. Although SNPs were selected on the basis of a likely involvement in the developmental process, it is not expected that all SNPs will be equally significant. Rather it is expected that significant effect sizes are concentrated in a relatively small number of SNP-phenotype combinations, since phenotypes are highly correlated across scalp electrodes for each sex-modality-age combination. Lack of concentration would suggest many of the large effect sizes (corresponding to low p-values) are the result of large random variation of effects across age and phenotype, and do not represent biologically stable phenomena. Concentration of effect is established by permutation tests which compare the realized distribution of effect sizes across SNP-phenotype combinations with the distribution of effect sizes of hypothetical random distributions of the realized effect sizes.
