APtools - Average Positive Predictive Values (AP) for Binary Outcomes and
Censored Event Times
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).