DelayedSurvFit: Survival functions with single-crossing constraints

In nchenderson/DelayedSurvFit: Constrained Survival Estimation with Delayed Treatment

Description

This computes nonparametric estimates of the survival curves in two treatment arms under the constraint that the two survival curves can cross at most one time.

Usage

DelayedSurvFit(times, events, trt, gamma=NULL, theta.fixed=NULL, max.times=100,
               inner.iter=50, final.iter=1000, verbose=TRUE)

Arguments

times	vector of follow-up times
events	vector of event indicators
trt	vector of treatment arm assignment
gamma	value of the gamma parameter
theta.fixed	optional fixed value of theta. Estimated if not provided.
max.times	tbf
inner.iter	tbf
final.iter	tbf
verbose	tbf

Value

an object of class surv.delay

Author(s)

Nicholas Henderson

Examples

set.seed(5172)

### Generate simulated survival data
n <- 400
cens.time <- 8
X1 <- rweibull(n, shape=2.3, scale=7)
X2 <- rweibull(n, shape=.8, scale=10)
Y1 <- pmin(X1, cens.time)
Y2 <- pmin(X2, cens.time)
e1 <- as.numeric(X1 < cens.time)
e2 <- as.numeric(X2 < cens.time)

times <- c(Y1, Y2)  ## follow-up times
events <- c(e1, e2) ## event indicators
trt <- rep(c(0,1), each=n)  ## treatment arm assignment

dsf.obj <- DelayedSurvFit(times, events, trt, gamma=-1, theta.fixed=5.8, max.times=100) 

km0 <- cumprod(1 - dsf.obj$nevents0/dsf.obj$nrisk0)
km1 <- cumprod(1 - dsf.obj$nevents1/dsf.obj$nrisk1)

plot(dsf.obj)
lines(dsf.obj$times, km0, type="s", lty=2)
lines(dsf.obj$times, km1, type="s", lty=2, col="red")

nchenderson/DelayedSurvFit documentation built on Jan. 21, 2021, 9:15 a.m.