timereg (version 1.9.3)

prep.comp.risk: Set up weights for delayed-entry competing risks data for comp.risk function

Description

Computes the weights of Geskus (2011) modified to the setting of the comp.risk function. The returned weights are \(1/(H(T_i)*G_c(min(T_i,tau)))\) and tau is the max of the times argument, here \(H\) is the estimator of the truncation distribution and \(G_c\) is the right censoring distribution.

Usage

prep.comp.risk(data, times = NULL, entrytime = NULL, time = "time",
  cause = "cause", cname = "cweight", tname = "tweight",
  strata = NULL, nocens.out = TRUE, cens.formula = NULL,
  cens.code = 0, prec.factor = 100, trunc.mintau = FALSE)

Arguments

data

data frame for comp.risk.

times

times for estimating equations.

entrytime

name of delayed entry variable, if not given computes right-censoring case.

time

name of survival time variable.

cause

name of cause indicator

cname

name of censoring weight.

tname

name of truncation weight.

strata

strata variable to obtain stratified weights.

nocens.out

returns only uncensored part of data-frame

cens.formula

censoring model formula for Cox models for the truncation and censoring model.

cens.code

code for censoring among causes.

prec.factor

precision factor, for ties between censoring/even times, truncation times/event times

trunc.mintau

specicies wether the truncation distribution is evaluated in death times or death times minimum max(times), FALSE makes the estimator equivalent to Kaplan-Meier (in the no covariate case).

Value

Returns an object. With the following arguments:

dataw

a data.frame with weights.

The function wants to make two new variables "weights" and "cw" so if these already are in the data frame it tries to add an "_" in the names.

References

Geskus (2011), Cause-Specific Cumulative Incidence Estimation and the Fine and Gray Model Under Both Left Truncation and Right Censoring, Biometrics (2011), pp 39-49.

Shen (2011), Proportional subdistribution hazards regression for left-truncated competing risks data, Journal of Nonparametric Statistics (2011), 23, 885-895

Examples

# NOT RUN {
data(bmt)
nn <- nrow(bmt)
entrytime <- rbinom(nn,1,0.5)*(bmt$time*runif(nn))
bmt$entrytime <- entrytime
times <- seq(5,70,by=1)

### adds weights to uncensored observations
bmtw <- prep.comp.risk(bmt,times=times,time="time",
		       entrytime="entrytime",cause="cause")

#########################################
### nonparametric estimates
#########################################
## {{{ 
### nonparametric estimates, right-censoring only 
out <- comp.risk(Event(time,cause)~+1,data=bmt,
		 cause=1,model="rcif2",
		 times=c(5,30,70),n.sim=0)
out$cum
### same as 
###out <- prodlim(Hist(time,cause)~+1,data=bmt)
###summary(out,cause="1",times=c(5,30,70))

### with truncation 
out <- comp.risk(Event(time,cause)~+1,data=bmtw,cause=1,
  model="rcif2",
  cens.weight=bmtw$cw,weights=bmtw$weights,times=c(5,30,70),
  n.sim=0)
out$cum
### same as
###out <- prodlim(Hist(entry=entrytime,time,cause)~+1,data=bmt)
###summary(out,cause="1",times=c(5,30,70))
## }}} 

#########################################
### Regression 
#########################################
## {{{ 
### with truncation correction
out <- comp.risk(Event(time,cause)~const(tcell)+const(platelet),data=bmtw,
 cause=1,cens.weight=bmtw$cw,
 weights=bmtw$weights,times=times,n.sim=0)
summary(out)

### with only righ-censoring, standard call
outn <- comp.risk(Event(time,cause)~const(tcell)+const(platelet),data=bmt,
	  cause=1,times=times,n.sim=0)
summary(outn)
## }}} 


# }

Run the code above in your browser using DataCamp Workspace