solvear.pchsurv: Penalized maximum likelihood estimation from the piecewise...

In obouaziz/pchsurv: Piecewise constant estimation of the hazard function for time-to-event data

Description

Given time to event data, a set of cuts and a penalty value, returns the penalized maximum likelihood estimator of the log-hazard using the adaptive-ridge penalty. The penalty is iteratively updated until convergence of the algorithm.

Usage

solvear.pchsurv(time, status, cuts, w = rep(1, length(cuts)), pen = 2,
  a = NULL, J = NULL, tol = 1e-07, itermax = 50, weights = rep(1,
  length(time)))

Arguments

time	the observed time data.
status	status of the time data. TRUE for true time and FALSE for censored time.
cuts	a sequence of cuts. Default to NULL which corresponds to the exponential model.
pen	the penalty value
a	the log-hazard value at the initialization step. Default to the unpenalized log-hazard estimator.
tol	the tolerance parameter for convergence of the algorithm. Default to 1e-7.
itermax	the maximum number of iterations. If the algorithm has not converged before itermax iterations then the algorithm exits the program.
weights	the weight values at the initialization step. Default to 1.
weights	an optional weight sequence. Default to 1.

Examples

n=400
cuts=c(20,40,50,70)
alpha=c(0,0.05,0.1,0.2,0.4)/10
time=rsurv(n,cuts,alpha) #generate true data from the pch model
censoring=runif(n,min=70,max=90)
time=pmin(time,censoring) #observed times
delta=time<censoring #gives 62% of observed data on average
result1=solvear.pchsurv(time,delta,cuts,pen=2)
result2=solvear.pchsurv(time,delta,cuts,pen=0)
result3=mlepchsurv(time,delta,cuts) #compare exp(result2) with exp(result3$a)!

obouaziz/pchsurv documentation built on Sept. 7, 2020, 11:03 a.m.