Figure 2

Diagnostic plots. The goodness-of-fit of the two estimators of G and densities of p-values for three data sets illustrating different scenarios in the performance of CGHpower. The data set of Douglas et al. shows A) a satisfactory goodness-of-fit following from B) a convex p-value density function. Mediocre operation is demonstrated with the data set of Postma et al. C) An inferior fit results from D) a p-value density which shows a slight increase for small values, but is not convex as expected. Nymark et al. represents failed execution. E) The disagreement between the G estimators is slightly more severe and the estimated power curve is a flat line (Figure 1). F) P-values exhibit even less density at low values than would be expected by chance. In such circumstances, it is recommended that data preprocessing be carried out before uploading and only the power calculations part be performed in CGHpower.