surrosurvROC {surrosurvROC}

Surrogate Survival ROC

Description

Nonparametric and semiparametric estimations of the time-dependent ROC curve for an incomplete failure time data with surrogate failure time endpoints

Usage

  surrosurvROC(DATA, method, pred.time, wt=NULL, span=NULL, b.rep=200)

Arguments

DATA

data frame, consisting of Marker: Predictior or marekr value; Survival time; Status: Event indicator (1: event; 0: censoring); STime: Surroagte survival Time; SStatus: Surrogate event indicator (1: event; 0: censoring)

method

"KNN"" for nonparametric model using nearest neighborhood kernel; "COX"" for semiparametric proportional hazard model

pred.time

Prediction time of the ROC curve

wt

Weight, such as inverse probablity weighting

span

Smoothing bandwidth parameter for KNN

b.rep

Number of bootstrap

Details

It provides a more efficient time-dependent ROC curve for an incomplete failure time data, when surrogate failure time endpoints are additionally observed for all subjects.

Author(s)

Yunro Chung [cre]

References

Yunro Chung and Yingye Zheng, Evaluating Prognostic Accuracy of Biomarkers for Incomplete and Right-Censored Data with Surrogate Outcome (in progress)

Examples

DATA=data.frame(
  Time=   c(1,2,5,3,9,NA,8,9,10,NA,NA,NA,6,4,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA),
  Status= c(1,1,0,0,1,NA,1,1,0, NA,NA,NA,0,0,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA),
  STime=  c(3,2,4,2,8,5,8,7,11,1,8,9,3,5,2,5,10,3,5,8,5,2,4,6,7),
  SStatus=c(1,0,1,0,1,1,1,0,0,1,1,1,1,0,1,1,0,0,1,0,1,0,1,0,0),
  Marker= c(1,5,1,2,3,1,2,3,4,5,9,8,5,7,3,4,2,5,3,4,7,5,9,3,8)
)

#COX at year 3
RES1=surrosurvROC(DATA, method="COX", pred.time=3)
print(RES1)

#KNN at year 3
nobs=sum(!is.na(DATA$Time))
span=0.25*nobs^(-0.20)
RES2=surrosurvROC(DATA, method="KNN",pred.time=3,span=span)
print(RES2)