Fig 3. Fold Increase in Number of False Positive Tests for Model2 vs. Model1.

Fig 3 displays the impact of mild departures in the null distribution of nominal p-values on the false positive rate. If one assumes the null p-value density is uniform (Model1; Gamma(1,1)) when in fact the null density is the given in Model2, Gamma(0.9334, 1.1144), using the significance levels as determined from Model1 will generate higher than expected false positive rates. The fold increase in this false positive inflation is plotted as a function of the significance level.