Re: Re: st: Comparison of Coefficients in Conditional Logistic regression

[Maarten Buis <maartenlbuis@gmail.com>]

On Mon, Jun 18, 2012 at 10:18 PM, A. Karvounidis <A.Karvounidis@uvt.nl> wrote:
> Thanks everyone for the help, all these papers and proposed models include an interaction variable in their method, but in my case there is NOT interaction variable and I control for the effect of my moderator by running again the regression with part of my initial sample( I have exclude from my initial sample companies with service activities in order to check only for the manufacture activity companies). So how can I compare the coefficients of two regressions (same type of regression)? Thats the question...

The answer is that you must re-estimate your model as a model with
interactions. You said before that you could not get that model to
converge. In that case I would first look at centering all variables
at some appropriate value within the range of the data before creating
the interactions. If that does not work, derive good starting values
from the separately estimated models(*), and use those. If that does
not work try various options in -help clogit##maximize_options-. And
if that does not work than your model is just inestimable.

-- Maarten

(*) The coefficients from the separate models are related but not the
same as the coefficients from a model with interactions. So you need
to do some computations before doing so.

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   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/