Statalist


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

Re: st: clogit data format


From   "Arne Risa Hole" <arnehole@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: clogit data format
Date   Thu, 20 Mar 2008 16:10:29 +0000

Hi Margaret,

As Nick suggests it is hard to say what the reason for the discrepancy
is without knowing more details about your application. Judging from
the example on the website you referred to, however, I think the
explanation might be that -clogit- can be used to estimate two very
different models: the fixed effects logit model (which can also be
estimated using -xtlogit, fe-) and McFadden's choice model (which can
also be estimated using the new -asclogit- command). In the example it
is shown how McFadden's model can be estimated using both -clogit- and
PHREG but from the description you gave it sounds like what you want
is a fixed effects logit. -clogit- is still appropriate but PHREG is
probably not, although I don't know this for certain.

As a side note it is slightly unfortunate in my opinion that Stata has
a single estimation command for these very different models (even
though they are, of course, identical from a computational point of
view). I suspect the addition of the -asclogit- command in Stata 10 is
an attempt to reduce the confusion that the -clogit- command has
caused as there is now a dedicated command for each model.

Hope this helps.

Arne


On 20/03/2008, Margaret R Grove <margaret.r.grove@dartmouth.edu> wrote:
> To clarify further where I think the problem may lie:
>
> PHREG output notes "Number of Observations Read   2300" and "Number of
> Observations Used  2000" (300 have missing values for the dependent
> variable)
>
> CLOGIT notes that 1710 observations (496 groups) were dropped because of
> all negative or all positive outcomes and our final number of
> observations is 290!
>
> With this I wonder if comparing the two methods makes sense and which
> method (PHREG or CLOGIT) is preferable (if any)?
>
> Thank you!
> Margaret
>
>
>
>
> Nick Cox wrote:
> > Be that as it may, the report of "very different answers" is pretty
> > difficult
> > to discuss without any details whatsoever.
> >
> > Nick
> > n.j.cox@durham.ac.uk
> >
> > Margaret R Grove
> >
> > I've seen examples of SAS PHREG and Stata clogit used on the same data
> > at the following web site:
> >
> > http://www.indiana.edu/~statmath/stat/all/cdvm/cdvm7.html
> >
> > Svend Juul wrote:
> >
> >
> >> Margaret wrote:
> >>
> >> I'm attempting to use clogit and have a very basic
> >> question about the data format. It seems like it
> >> should be obvious but I'm getting very different
> >> answers than a colleague using SAS PHREG.
> >> ...
> >>
> >> ====================================================
> >>
> >> Isn't SAS PHREG about proportional hazards regression
> >> (Cox regression)? Stata's -clogit- performs conditional
> >> logistic regression; in my understanding this is quite
> >> a different analysis.
> >>
> >
> > *
> > *   For searches and help try:
> > *   http://www.stata.com/support/faqs/res/findit.html
> > *   http://www.stata.com/support/statalist/faq
> > *   http://www.ats.ucla.edu/stat/stata/
> >
>
> --
> __________________________
> Margaret R. Grove, MS
> Research Associate/Analyst, TDI
> 1 Medical Center Drive, Room 570
> Lebanon, NH 03756
> Tel: 603-653-3560
> Fax: 603-653-3558
>
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index