Re: st: RE: Survival analysis - individual survival functions

[Steven Samuels <sjsamuels@gmail.com>]

-

You are very welcome, Paul, but I think that using the post-estimation  
facilities of -stpm2-, as outlined by Paul Lambert, might be better,  
as well as easier.  -stpm2- has the advantage of being parametric, but  
doesn't require that you pick (and defend!) a parametric distribution.

In any case, with -streg- you'll need  -nlcom- to produce  estimates  
and CIs on a suitable scale and then transform back to the probability  
scale. With -stpm2- it would be simpler to bootstrap, given the ease  
of producing S(t)/S(t_0).


Steve
sjsamuels@gmail.com



On Dec 16, 2010, at 3:31 PM, Seed, Paul wrote:

Thanks Steve.
I had been hoping to avoid doing all the algebra (or at
least get some pre-calculated values to check my algebra
against).

It6 seems I'm out of luck.

Date: Wed, 15 Dec 2010 20:09:30 -0500
From: Steven Samuels <sjsamuels@gmail.com>
Subject: Re: st: Survival analysis - individual survival functions

Paul,

 -predict- gives values at the observations after every regression
command. To get it for arbitrary values of the inputs, you have to do
it yourself.   S(t| t_0) = S(t)/S(t_0), and the formulas for S(.) are
on page 358 of the Stata Manual (V 11).  You can get the linear
predictor for any values of the X's with  -margins-  or -adjust-
( Stata<11).  Add the appropriate auxiliary parameters (returned
results e(theta), e(aux_p), e(gamma), e(sigma), e(kappa)); your t and
t_0; then plug  everything into the formulas.

Steve

Steven J. Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/