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).

install.packages("APtools")

Try the APtools package in your browser