timereg (version 1.8.6)

predict.timereg: Predictions for Survival and Competings Risks Regression for timereg

Description

Make predictions based on the survival models (Aalen and Cox-Aalen) and the competing risks models for the cumulative incidence function (comp.risk). Computes confidence intervals and confidence bands based on resampling.

Usage

## S3 method for class 'timereg':
predict(object,newdata=NULL,X=NULL,times=NULL,Z=NULL,
n.sim=500,uniform=TRUE,se=TRUE,alpha=0.05,resample.iid=0,...)

Arguments

object

an object belonging to one of the following classes: comprisk, aalen or cox.aalen

newdata

specifies the data at which the predictions are wanted.

X

alternative to newdata, specifies the nonparametric components for predictions.

Z

alternative to newdata, specifies the parametric components of the model for predictions.

times

times in which predictions are computed, default is all time-points for baseline

n.sim

number of simulations in resampling.

uniform

computes resampling based uniform confidence bands.

se

computes pointwise standard errors

alpha

specificies the significance levelwhich cause we consider.

resample.iid

set to 1 to return iid decomposition of estimates, 3-dim matrix (predictions x times x subjects)

...

unused arguments - for S3 compatability

Value

• time
  vector of time points where the predictions are computed.
• unif.band
  resampling based constant to construct 95% uniform confidence bands.
• model
  specifies what model that was fitted.
• alpha
  specifies the significance level for the confidence intervals. This relates directly to the constant given in unif.band.
• newdata
  specifies the newdata given in the call.
• RR
  gives relative risk terms for Cox-type models.
• call
  gives call for predict funtion.
• initial.call
  gives call for underlying object used for predictions.
• P1
  gives cumulative inicidence predictions for competing risks models. Predictions given in matrix form with different subjects in different rows.
• S0
  gives survival predictions for survival models. Predictions given in matrix form with different subjects in different rows.
• se.P1
  pointwise standard errors for predictions of P1.
• se.S0
  pointwise standard errors for predictions of S0.

References

Scheike, Zhang and Gerds (2008), Predicting cumulative incidence probability by direct binomial regression, Biometrika, 95, 205-220.

Scheike and Zhang (2007), Flexible competing risks regression modelling and goodness of fit, LIDA, 14, 464-483 .

Martinussen and Scheike (2006), Dynamic regression models for survival data, Springer.

Examples

library(timereg)
data(bmt); 

add<-comp.risk(Event(time,cause)~platelet+age+tcell,data=bmt,cause=1)

ndata<-data.frame(platelet=c(1,0,0),age=c(0,1,0),tcell=c(0,0,1))
out<-predict(add,newdata=ndata,uniform=1,n.sim=1000)
par(mfrow=c(2,2))
plot(out,multiple=0,uniform=1,col=1:3,lty=1,se=1)
# see comp.risk for further examples. 

add<-comp.risk(Event(time,cause)~factor(tcell),data=bmt,cause=1)
summary(add)
out<-predict(add,newdata=ndata,uniform=1,n.sim=1000)
plot(out,multiple=1,uniform=1,col=1:3,lty=1,se=1)

## SURVIVAL predictions aalen function
data(sTRACE)
out<-aalen(Surv(time,status==9)~sex+ diabetes+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0,0,0),rep(1,5)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

out<-aalen(Surv(time,status==9)~const(age)+const(sex)+
const(diabetes)+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0),c(1,1,0)),
Z=rbind(c(55,0,1),c(60,1,1)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

pout<-predict(out,uniform=0,se=0,newdata=sTRACE[1:10,]) 
plot(pout,multiple=1,se=0,uniform=0)

out<-cox.aalen(Surv(time,status==9)~prop(age)+prop(sex)+
prop(diabetes)+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0),c(1,1,0)),Z=rbind(c(55,0,1),c(60,1,1)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

pout<-predict(out,uniform=0,se=0,newdata=sTRACE[1:10,]) 
plot(pout,multiple=1,se=0,uniform=0)