In elong0527/simuCPP: Simulate Concordance Probability, AUC and Time-Dependent AUC Based on Copula

library(simuCCP)
library(mvtnorm)

Simulation

We can use simuCCP::simuCCP to simulate data with pre-defined C-index and time dependent AUC.

• metrics define different preditive metrics
• family define the copula family

Single Marker

For a Gaussian copula (family = 1), we can simulate a dataset for C-index = 0.7.

db1 <- simuCCP(N = 200, metric = "Cind", value = 0.7, family = 1)
head(db1$data)

Two Markers

For two markers with Gaussian and Clayton copulas (family = c(1,3)), we can simulate data for C-index both equal to 0.7. The conditional copula between two markers is Gumbel (fam2 = 4) with parameter 1 (par2 = 1)

db2 <- simuCCP(300, metric = "Cind", value = 0.7, family = c(1,3), fam2 = 4, par2 = 1)
head(db2$data)

Define parameters and predictive metrics

There are a set of functions to calculate parameter to predictive metric and vice versa.

• C-index: simuCCP::BiCopCind2Par and simuCCP::BiCopPar2Cind
• AUC: simuCCP::BiCopAUC2Par and simuCCP::BiCopPar2AUC

For example, we can find a clayton copula parameter for C-index equal to 0.75.

simuCCP::BiCopCind2Par(family = 3, Cind = 0.75)

elong0527/simuCPP documentation built on March 29, 2021, 10:03 a.m.