[18930] in s-news-athena
[S] Question on estimation of parametric survival curve
daemon@ATHENA.MIT.EDU (Suraj Rao)
Thu Aug 12 18:31:58 1999
Message-Id: <199908122224.RAA18791@stat.spdc.ti.com>
To: s-news@wubios.wustl.edu
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Thu, 12 Aug 1999 17:24:39 -0500
From: Suraj Rao <rao@stat.spdc.ti.com>
Dear Splusers,
Have a question on parameter estimation for survival analysis.
Assume the lifetime table below.
start end deaths withdraws
0 3 7 0
3 8 2 0
8 12 3 0
12 24 1 0
24 24 0 7696
One can use the KMG algorithm to build a non-parametric estimate of the
survival curve and the standard error.
I want to check this against assumption of, say, lognormal and Weibull
distribution. Are there any built-in functions that will allow one to obtain
Maximum likelihood estimates (and SE) for a parametric distribution.
I am hoping survreg can be coerced into doing the estimation but I do not know
how (manuals did not help). I would appreciate any comments on possibilities
or even validity
of such an exercise.
I would also like comments on the validity of the following procedure:
1) Linearize the failure probabilities obtained from the KMG algorithm
on a log normal plot.
2) Perform a least square fit against log(time).
3) Obtain slope, intercept and their standard errors (SE)
4) Estimate log-normal distribution parameters and SEs using delta method.
A more refined strategy would also include SE information from KMG algorithm
through the following steps:
1a) Apart from 1) above, use delta method to obtain SE in the lognormal scale
2a) Instead of 2) above, perform WEIGHTED least squares with inverse
SE as the weights
Continue with 3 and 4
My fear is that the above approach is ad hoc and also relies heavily on
repeated use of Delta method. Any comments on this procedure would also be
appreciated.
Regards,
Suraj
-------------------------------------------------------------------
Suraj Rao | E-mail: rao@spdc.ti.com
Texas Instruments Incorporated | Phone : (214) 927 3064
13536 North Central Expressway, MS 3735 |
Dallas, TX 75243 |
---------------------------------------------------------------------
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
