# Predicting survival probabilities Function to extract survival probability predictions from various modeling approaches. The most prominent one is the Cox regression model which can be fitted for example with `coxph' and with `cph'. ```r ## S3 method for class 'aalen' predictSurvProb(object,newdata,times,...) ## S3 method for class 'riskRegression' predictSurvProb(object,newdata,times,...) ## S3 method for class 'cox.aalen' predictSurvProb(object,newdata,times,...) ## S3 method for class 'cph' predictSurvProb(object,newdata,times,...) ## S3 method for class 'coxph' predictSurvProb(object,newdata,times,...) ## S3 method for class 'matrix' predictSurvProb(object,newdata,times,...) ## S3 method for class 'selectCox' predictSurvProb(object,newdata,times,...) ## S3 method for class 'pecCforest' predictSurvProb(object,newdata,times,...) ## S3 method for class 'prodlim' predictSurvProb(object,newdata,times,...) ## S3 method for class 'psm' predictSurvProb(object,newdata,times,...) ## S3 method for class 'survfit' predictSurvProb(object,newdata,times,...) ## S3 method for class 'pecRpart' predictSurvProb(object,newdata,times,...) #' \method{predictSurvProb}{pecCtree}(object,newdata,times,...) ``` ## Arguments - `object`: A fitted model from which to extract predicted survival probabilities - `newdata`: A data frame containing predictor variable combinations for which to compute predicted survival probabilities. - `times`: A vector of times in the range of the response variable, e.g. times when the response is a survival object, at which to return the survival probabilities. - ``...``: Additional arguments that are passed on to the current method. ## Returns A matrix with as many rows as `NROW(newdata)` and as many columns as `length(times)`. Each entry should be a probability and in rows the values should be decreasing. ## Details The function predictSurvProb is a generic function that means it invokes specifically designed functions depending on the 'class' of the first argument. The function `pec` requires survival probabilities for each row in newdata at requested times. These probabilities are extracted from a fitted model of class `CLASS` with the function `predictSurvProb.CLASS`. Currently there are `predictSurvProb` methods for objects of class cph (library rms), coxph (library survival), aalen (library timereg), cox.aalen (library timereg), rpart (library rpart), product.limit (library prodlim), survfit (library survival), psm (library rms) ## Note In order to assess the predictive performance of a new survival model a specific `predictSurvProb` S3 method has to be written. For examples, see the bodies of the existing methods. The performance of the assessment procedure, in particular for resampling where the model is repeatedly evaluated, will be improved by supressing in the call to the model all the computations that are not needed for probability prediction. For example, `se.fit=FALSE` can be set in the call to `cph`. ## Examples ```r # generate some survival data library(prodlim) set.seed(100) d <- SimSurv(100) # then fit a Cox model library(survival) library(rms) coxmodel <- cph(Surv(time,status)~X1+X2,data=d,surv=TRUE) # Extract predicted survival probabilities # at selected time-points: ttt <- quantile(d$time) # for selected predictor values: ndat <- data.frame(X1=c(0.25,0.25,-0.05,0.05),X2=c(0,1,0,1)) # as follows predictSurvProb(coxmodel,newdata=ndat,times=ttt) # stratified cox model sfit <- coxph(Surv(time,status)~strata(X1)+X2,data=d,,x=TRUE,y=TRUE) predictSurvProb(sfit,newdata=d[1:3,],times=c(1,3,5,10)) ## simulate some learning and some validation data learndat <- SimSurv(100) valdat <- SimSurv(100) ## use the learning data to fit a Cox model library(survival) fitCox <- coxph(Surv(time,status)~X1+X2,data=learndat,x=TRUE,y=TRUE) ## suppose we want to predict the survival probabilities for all patients ## in the validation data at the following time points: ## 0, 12, 24, 36, 48, 60 psurv <- predictSurvProb(fitCox,newdata=valdat,times=seq(0,60,12)) ## This is a matrix with survival probabilities ## one column for each of the 5 time points ## one row for each validation set individual ## Cox with ridge option f1 <- coxph(Surv(time,status)~X1+X2,data=learndat,x=TRUE,y=TRUE) f2 <- coxph(Surv(time,status)~ridge(X1)+ridge(X2),data=learndat,x=TRUE,y=TRUE) plot(predictSurvProb(f1,newdata=valdat,times=10), pec:::predictSurvProb.coxph(f2,newdata=valdat,times=10), xlim=c(0,1), ylim=c(0,1), xlab="Unpenalized predicted survival chance at 10", ylab="Ridge predicted survival chance at 10") ``` ## References Ulla B. Mogensen, Hemant Ishwaran, Thomas A. Gerds (2012). Evaluating Random Forests for Survival Analysis Using Prediction Error Curves. Journal of Statistical Software, 50(11), 1-23. DOI 10.18637/jss.v050.i11 ## See Also `predict`,`survfit` ## Author(s) Thomas A. Gerds tag@biostat.ku.dk

predictSurvProb function