popEpi (version 0.2.1)

rpcurve: Marginal piecewise parametric relative survival curve

Description

Fit a marginal relative survival curve based on a relpois fit

Usage

rpcurve(object = NULL, conf.int = 0.95)

Arguments

object

a relpois object

conf.int

confidence interval level; e.g. 0.95 for 95 % confidence intervals (region)

Details

Currently only estimates a marginal curve, i.e. the average of all possible individual curves. The confidence intervals are based on an assumption of asymptotic normalcy at the cumulative hazard level. Only supported when the reserved FOT variable was used in relpois. Computes a curve for each unique combination of covariates (e.g. 4 sets) and returns a weighted average curve based on the counts of subjects for each combination (e.g. 1000, 125, 50, 25 respectively). Fairly fast when only factor variables have been used, otherwise go get a cup of coffee. If delayed entry is present in data due to period analysis limiting, the marginal curve is constructed only for those whose follow-up started in the respective period.

Examples

## use the simulated rectal cancer cohort
sr <- copy(sire)
ab <- c(0,45,55,65,70,Inf)
sr$agegr <- cut(sr$dg_age, breaks = ab, right = FALSE)

BL <- list(fot= seq(0,10,1/12))
x <- lexpand(sr, breaks=BL, pophaz=popmort, status=status)

rpm <- relpois(x, formula = lex.Xst %in% 1:2 ~ -1+ FOT + agegr, fot.breaks=c(0,0.25,0.5,1:8,10))
pmc <- rpcurve(rpm)

## compare with non-parametric estimates
st <- survtab(x,relsurv.method = "e2", agegr.w.breaks=c(0,45,55,65,75,Inf))

plot(I(c(0.5,1))~I(c(0,10)), type="n", xlab="years", ylab="relative survival")
matlines(y = st[, list(r.e2.as, r.e2.as.lo, r.e2.as.hi)], x = st$Tstop, col="blue", lty=c(1,2,2))
matlines(y = pmc[, list(est, lo, hi)], x = pmc$Tstop, col="red", lty=c(1,2,2))