FGR: Formula wrapper for crr from cmprsk

In riskRegression: Risk Regression Models and Prediction Scores for Survival Analysis with Competing Risks

Formula wrapper for crr from cmprsk

Description

Formula interface for Fine-Gray regression competing risk models.

Usage

FGR(formula, data, cause = 1, y = TRUE, ...)

Arguments

formula	A formula whose left hand side is a Hist object – see Hist. The right hand side specifies (a linear combination of) the covariates. See examples below.
data	A data.frame in which all the variables of formula can be interpreted.
cause	The failure type of interest. Defaults to 1.
y	logical value: if TRUE, the response vector is returned in component response.
...	...

Details

Formula interface for the function crr from the cmprsk package.

The function crr allows to multiply some covariates by time before they enter the linear predictor. This can be achieved with the formula interface, however, the code becomes a little cumbersome. See the examples. Note that FGR does not allow for delayed entry (left-truncation). The assumed value for indicating censored observations in the event variable is 0. The function Hist has an argument cens.code which can change this (if you do not want to change the event variable).

Value

See crr.

Author(s)

Thomas Alexander Gerds [e-mail]

References

Gerds, TA and Scheike, T and Andersen, PK (2011) Absolute risk regression for competing risks: interpretation, link functions and prediction Research report 11/7. Department of Biostatistics, University of Copenhagen

See Also

riskRegression

Examples

library(prodlim)
library(survival)
library(cmprsk)
library(lava)
d <- prodlim::SimCompRisk(100)
f1 <- FGR(Hist(time,cause)~X1+X2,data=d)
print(f1)

## crr allows that some covariates are multiplied by
## a function of time (see argument tf of crr)
## by FGR uses the identity matrix
f2 <- FGR(Hist(time,cause)~cov2(X1)+X2,data=d)
print(f2)

## same thing, but more explicit:
f3 <- FGR(Hist(time,cause)~cov2(X1)+cov1(X2),data=d)
print(f3)

## both variables can enter cov2:
f4 <- FGR(Hist(time,cause)~cov2(X1)+cov2(X2),data=d)
print(f4)

## change the function of time
qFun <- function(x){x^2}
noFun <- function(x){x}
sqFun <- function(x){x^0.5}

## multiply X1 by time^2 and X2 by time:
f5 <- FGR(Hist(time,cause)~cov2(X1,tf=qFun)+cov2(X2),data=d)
print(f5)
print(f5$crrFit)
## same results as crr
with(d,crr(ftime=time,
           fstatus=cause,
           cov2=d[,c("X1","X2")],
           tf=function(time){cbind(qFun(time),time)}))

## still same result, but more explicit
f5a <- FGR(Hist(time,cause)~cov2(X1,tf=qFun)+cov2(X2,tf=noFun),data=d)
f5a$crrFit

## multiply X1 by time^2 and X2 by sqrt(time)
f5b <- FGR(Hist(time,cause)~cov2(X1,tf=qFun)+cov2(X2,tf=sqFun),data=d,cause=1)

## additional arguments for crr
f6<- FGR(Hist(time,cause)~X1+X2,data=d, cause=1,gtol=1e-5)
f6
f6a<- FGR(Hist(time,cause)~X1+X2,data=d, cause=1,gtol=0.1)
f6a

riskRegression documentation built on Sept. 8, 2023, 6:12 p.m.