Average Positive Predictive Values (AP) for Binary Outcomes and
Censored Event Times
Description
We provide tools to estimate two prediction accuracy metrics,
the average positive predictive values (AP) as well as the well-known AUC
(the area under the receiver operator characteristic curve) for risk scores.
The outcome of interest is either binary or censored event time.
Note that for censored event time, our functions' estimates, the AP and the
AUC, are time-dependent for pre-specified time interval(s). A function that
compares the APs of two risk scores/markers is also included. Optional
outputs include positive predictive values and true positive fractions at
the specified marker cut-off values, and a plot of the time-dependent AP
versus time (available for event time data).