recurrent.marginal.coxmean {timereg} | R Documentation |
Fitting two Cox models for death and recurent events these are combined to prducte the estimator
\int_0^t S(u|x=0) dR(u|x=0)
the mean number of recurrent events, here
S(u|x=0)
is the probability of survival, and
dR(u|x=0)
is the probability of an event among survivors. For now the estimator is based on the two-baselines so
x=0
, but covariates can be rescaled to look at different x's and extensions possible.
recurrent.marginal.coxmean(recurrent, death)
recurrent |
aalen model for recurrent events |
death |
cox.aalen (cox) model for death events |
IID versions along the lines of Ghosh & Lin (2000) variance. See also mets package for quick version of this for large data. IID versions used for Ghosh & Lin (2000) variance. See also mets package for quick version of this for large data mets:::recurrent.marginal, these two version should give the same when there are now ties.
Thomas Scheike
Ghosh and Lin (2002) Nonparametric Analysis of Recurrent events and death, Biometrics, 554–562.
### do not test because iid slow and uses data from mets library(mets) data(base1cumhaz) data(base4cumhaz) data(drcumhaz) dr <- drcumhaz base1 <- base1cumhaz base4 <- base4cumhaz rr <- simRecurrent(100,base1,death.cumhaz=dr) rr$x <- rnorm(nrow(rr)) rr$strata <- floor((rr$id-0.01)/50) drename(rr) <- start+stop~entry+time ar <- cox.aalen(Surv(start,stop,status)~+1+prop(x)+cluster(id),data=rr, resample.iid=1,,max.clust=NULL,max.timepoint.sim=NULL) ad <- cox.aalen(Surv(start,stop,death)~+1+prop(x)+cluster(id),data=rr, resample.iid=1,,max.clust=NULL,max.timepoint.sim=NULL) mm <- recurrent.marginal.coxmean(ar,ad) with(mm,plot(times,mu,type="s")) with(mm,lines(times,mu+1.96*se.mu,type="s",lty=2)) with(mm,lines(times,mu-1.96*se.mu,type="s",lty=2))