[6446] in s-news-athena
agreg & phreg
daemon@ATHENA.MIT.EDU (Martin Hellmich)
Wed Feb 15 03:42:43 1995
Date: Wed, 15 Feb 95 09:11:43 +0100
From: hellmic@medsun01.uni-muenster.de (Martin Hellmich)
To: s-news@utstat.toronto.edu
Hi all,
is it possible to reproduce the analysis of Kalbfleisch & Prentice with phreg of
SAS? I DID FAIL to get the same coefficients as in their book. With agreg of
S-PLUS I had no trouble to do so. Well, I mailed Terry Therneau whether he know
something about it. He pointed out that
> because of the interaction terms, one of the
> coefficients in the K&P model is essentially arbitrary (you will notice that
> in the book, no standard errors are shown for this term). This coef is
> very dependent on the coding used, e.g., if "age" were replaced by
> "xage = age+10" the value will change.
Maybe you have a different idea or explanation?
|||
|||||
Martin Hellmich |||||||
Institute of Medical Informatics and Biomathematics |||||||
University of Muenster |||||||
Domagkstr. 9 Tel.: 0251/83-5276 | | | |
D-48129 Muenster Fax: 0251/83-5277
Germany E-mail: hellmic@medsun01.uni-muenster.de
