[R] Turnbull estimate

[Martyn Plummer]

On Fri, 2004-01-23 at 01:33, Denise wrote:
> Hi everyone,
> 
> I've been looking for a function that calculates the Turnbull estimate
> for left, right and interval censored data. None of the data that I am
> using has exact failure times. The only function I seem to find can handle
> data that has no left censoring. It seems that I will have to program
> this myself. In your opinion, is it advisable to start with the
> function, kaplanMeier, in Splus and try to get it compatible with R?
> I'm not sure if this is possible but I'm willing to try.
> 
> Any advice would be greatly appreciated!

A couple of years ago, I wrote some functions for analyzing interval
censored survival data. I put them in a package for convenience, but
they are by no means user friendly. I put a copy for you on

http://www-fis.iarc.fr/~martyn/software/interval/

Your mileage may vary. Quite possibly down to zero. 

Here is an example that plots the non-parametric maximum likelihood
estimate of the survival curve for the famous cosmesis data

package(interval)
data(cosmesis)
attach(cosmesis)
plot.surv(start, end, treat)

I hope this helps. Tidying it up and using the interface provided by
Terry Therneau's survival package are on my TODO list, but a long way
down.

Martyn