s-news
[Top] [All Lists]

Re: How to Use COXPH OBJECT

To: Jiahong Xu <JiahongXu@westat.com>
Subject: Re: How to Use COXPH OBJECT
From: Frank E Harrell Jr <fharrell@virginia.edu>
Date: Mon, 21 Apr 2003 11:32:59 -0400
Cc: s-news@lists.biostat.wustl.edu, jonasellenberg@westat.com
In-reply-to: <9B501B3774931C469BCCCC021BE53722693D22@remailnt2-re01.westat.com>
Organization: University of Virginia
References: <9B501B3774931C469BCCCC021BE53722693D22@remailnt2-re01.westat.com>
On Mon, 21 Apr 2003 11:10:43 -0400
Jiahong Xu <JiahongXu@westat.com> wrote:

> Dear Colleague:
>  
> I am using S+2000 (windows) and would like to know how to use COX model
> coefficients and statistics, without the original data set to estimate the
> baseline survival for a set of predictors including time-dependent
> covariates. 
>  
> The problem is that I don't have my raw dataset in S+ format, although I do
> have  the model coefficients as well as related model fit statistics
> generated from SAS. I was advised by S-plus support group that a COXPH
> OBJECT could do this job but needs some sophisticated programming using
> COEFFICIENTS, VAR, NAIVE.VAR, LOGLIK, SCORE, RSCORE, WALD.TEST, ITER, etc.
> Converting the original dataset from SAS to S+ is not possible.  
>  
> I would highly appreciate it if someone can give me some help  on how to use
> this very useful tool only knowing the model statistics and not having the
> original dataset.
>  
> Thank you very much in advance! 
>  
> Jiahong 
>   
> 
It would be more trouble to do this than to import the SAS dataset (using the 
S-Plus menus or the sas.get function in the Hmisc library) and re-fit the model 
using coxph.  Also take a look at the cph function in the Design library.  cph 
makes it easier to get certain predicted values in some cases.  For one thing, 
cph (which calls coxph routines) can make it easy to draw a nomogram depicting 
the model.

---
Frank E Harrell Jr              Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine  http://hesweb1.med.virginia.edu/biostat

<Prev in Thread] Current Thread [Next in Thread>