Figure 4

Mean predictive accuracy stratified by type of SNP. Each data point represents the mean of 100 trials. 95% confidence intervals are shown for the coding non-synonymous SNPs. Confidence interval widths are of comparable size for the other SNP categories. Predictive accuracy is very similar across all types of SNPs, although the small differences in mean predictive accuracy are statistically significant because of the large number of SNPs used in the analysis. For each n (where n is the number of randomly selected SNPs used to build the classifiers), one-way ANOVA was used to test the null hypothesis that means across SNP categories were equal. P-values were less than .01 for all tests where n>5, and p = .07 for n = 5.