Figure 6

Reasons for different performances.

(a) The positive area of different models. Using lower cutoff to cover the HQD, the positive area of model P will cover part of LQD, but that of OScal_B will not. (b) The average PPV of different models and that of different districts in the d-R plane.