Conditional Random Fields for Fast, Large-Scale Genome-Wide Association Studies
Figure 5
Quantile-quantile (QQ) plots for the GOLDN data.
QQ plots of model negative p-value statistics obtained from our method as a function of expected negative p-values under the null hypothesis for synthetic GOLDN data with (a) and real data (b). For comparison, negative p-value statistics obtained from a logistic regression that does not account for confounding factors and from Eigenstrat [4] are shown for the synthetic data with (c,e) and real data (d,f). Dotted red lines indicate 95% confidence bounds.