Chunk #41 — Statistical Methodology for the Analysis of Association Trajectories — Identification of significant SNPs — Significant SNPs identified by concentration of effect sizes
The permutation tests are scale-free, and their use to identify important SNPs depends upon the absolute values of the effect sizes themselves. Thus some effect size thresholding is necessary to make this method useful for identifying significant SNPs; conversely, if all effect sizes are significantly large, then the method is superfluous. The averaging procedure serves to reduce the variance of the aggregated data, and combined with information from bootstrapping, ensures the reliability of the effect size estimates, whose importance can be gauged from the results of other genetic studies. We note that effect sizes of the major allele of significant SNPs typical of GWAS results in Kang et al. (2012) are of absolute value between 0.1 and 0.2. In the particular distribution of effect sizes obtained in this study, the method confirmed, that except for a single SNP, a threshold of .15 in effect size would correspond to a threshold of 10−4 in p-value, as shown in the top panel of Figure 7.
