Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Testing -stcox- proportionality under -micombine-

From   Maarten buis <>
Subject   Re: st: Testing -stcox- proportionality under -micombine-
Date   Mon, 14 Aug 2006 21:00:31 +0100 (BST)

--- Andrew Clarkwest <> wrote:
> I have been using -ice- to perform multiple imputation of missing
> data, and I've been analyzing the subsequently created data using
> -micombine- with -stcox-. For example:
>        micombine stcox X1 X2 X3, sch(scha*)
> A problem that I've run up against is that I can't test the
> proportionality assumption in those Cox PH models. Specifically, the
> Schoenfeld and scaled Schoenfeld options can't be used under
> -micombine- because it's running separate models for each imputed
> dataset created by -ice- (in my case 5 of them). Once Stata creates
> the variables to store the Schoenfeld residuals for the first imputed
> dataset, it tries to do so again for the second, but using the same
> variable names as for the first, so I get an error message saying
> "scha1 already defined". I can't run the proportionality test (-estat
> phtest-) without the Schoenfeld residuals and, as of yet, I'm unable
> to save the residuals while using -micombine-.


As a "less than formal" test I would manually estimate a -stcox model-
and perform the proportionality test in each sample, and use the mean
of the different test statistics as your test statistic. This is
reasonably in concordance with MI practice, though not completely. This
approach is based on normality of the test statistic and if I remember
correctly the result of the proportionality test is Chi square with
little degrees of freedom.


Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

The all-new Yahoo! Mail goes wherever you go - free your email address from your Internet provider.
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index