pseudo (version 1.4.3)

pseudoyl: Pseudo-observations for the expected number of years lost

Description

Computes pseudo-observations for modeling using the number of years lost.

Usage

pseudoyl(time,event, tmax)

Arguments

time

the follow up time.

event

the cause indicator, use 0 as censoring code and integers to name the other causes.

tmax

the maximum cut-off point time = the upper limit of the integral of the cumulative incidence function. If missing or larger than the maximum follow up time, it is replaced by the maximum follow up time.

Value

A list containing the following objects:

cause

The ordered codes for different causes.

pseudo

A list of vectors- a vector for each of the causes, ordered by codes. Each value of a vector belongs to one individual (ordered as in the original data set).

Details

The function calculates the pseudo-observations for the expected number of years lost for each individual. The pseudo-observations can be used for fitting a regression model with a generalized estimating equation. No missing values in either time or event vector are allowed.

References

Andersen P.K.: "A note on the decomposition of number of life years lost according to causes of death." Research report, Department of Biostatistics, University of Copenhagen, 2012 (2)

See Also

pseudoci, pseudomean, pseudosurv

Examples

# NOT RUN {
library(KMsurv)
data(bmt)
bmt$icr <- bmt$d1 +  bmt$d3


#compute the pseudo-observations:
pseudo = pseudoyl(time=bmt$t2, event=bmt$icr,tmax=2000)

#arrange the data - use pseudo observations for cause 2
a <- cbind(bmt,pseudo = pseudo$pseudo[[2]],id=1:nrow(bmt))

#fit a regression model for cause 2

library(geepack)
summary(fit <- geese(pseudo ~ z1 + as.factor(z8) + as.factor(group),
	data = a, id=id, jack = TRUE, family=gaussian, 
	corstr="independence", scale.fix=FALSE))


#rearrange the output
round(cbind(mean = fit$beta,SD = sqrt(diag(fit$vbeta.ajs)),
	Z = fit$beta/sqrt(diag(fit$vbeta.ajs)),	PVal =
	2-2*pnorm(abs(fit$beta/sqrt(diag(fit$vbeta.ajs))))),4)
# }