Re: st: RE: Order of variables in stcox model

[Brigham Whitman <brighamj@gmail.com>]

I do get the same results when I run the same code twice.  Even when I
clear out the data and re-enter everything, I still get identical
results using the same code.  But these results change when I change
the order of the variables.  I am not changing the data or stset
between models.

Perhaps I am over-parameterizing the model, because I can remove one
variable/ stratum at a time until I get to a point (7 variables and 2
strata) where the results don't change when I change the order of the
variables.  Though I don't know why Stata would run the model
differently...

-Brigham




On Fri, Aug 12, 2011 at 1:35 PM, Sarah Edgington <sedging@ucla.edu> wrote:
> .
> The ordering of your variables should not matter.  Are you making any
> changes to the data between models?  Is the data stset the same for every
> model?  If you run the same code more than once do you get the same results?
> My suspicion is that the answer to this last question is no and the problem
> actually has to do with the sort order of the data rather than the order of
> the variables in the model.  I have in the past run into problems with this
> kind of model when the data is not sorted the same across runs.  If this is
> the problem you may notice that if you run the same dofile multiple times
> (without changing the saved dataset it opens) that you get different
> results.
> -Sarah
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Brigham Whitman
> Sent: Friday, August 12, 2011 9:52 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Order of variables in stcox model
>
> Hello.  I am building a model using stcox in Stata 9.2.
>
> My question is: How should I order the variables that I put in the model?  I
> get different results depending on how the variables are ordered.  For
> example:
>
> xi: stcox  velocity roads snow i.season, nohr strata(sex  age) robust
>
> will produce different results than:
>
> xi: stcox  roads snow velocity i.season, nohr strata(sex  age) robust
>
> To be completely transparent: with my data, sometimes the order of the
> variables does not matter for smaller models (less than 4 variables).
> But with larger models (more than 6 variables), the results do change when
> the order of the variables is changed.
>
> What is the importance of the order of the variables (in this case, roads,
> snow, velocity, and i.season) in the model?
>
> Thank you,
>
> Brigham Whitman
> *
> *   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/
>

*
*   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/