Re: st: Comparison of Coefficients in Conditional Logistic regression

[Nick Cox <njcoxstata@gmail.com>]

Does any of these references address the question asked?

It seems to me that it is futile to seek a significance level or
P-value for the differences between Anestis' models. They appear to
result from extensive interaction with the data and so all innocence
has been sacrificed. In fact it sounds to me that the bigger problem
is justifying ad hoc fits (of non-convergent models?) to different
subsets. No matter who is in judgement -- paper reviewers, thesis
committee, whatever -- they are likely to give you a hard time on
that.

Nick

On Sat, Jun 16, 2012 at 7:40 PM, Cameron McIntosh <cnm100@hotmail.com> wrote:

> You need to be careful when you compare logistic regression coefficients. You should read the following:
>
> Fielding, A. (2004). Scaling for residual variance components of ordered category responses in generalised linear mixed multilevel models. Quality and Quantity, 38, 425-433.
>
> Bauer, D.J. (2009). A note on comparing the estimates of models for cluster-correlated or longitudinal data with binary or ordinal outcomes. Psychometrika, 74(1), 97-105.
>
> Allison, P.D. (1999). Comparing logit and probit coefficients across groups. Sociological Methods & Research, 28(2), 186-208.
>
> Allison, P.D. (1987). Introducing a Disturbance into Logit and Probit Regression Models. Sociological Methods & Research, 15(4), 355-374.
>
> Karlson, A.K.B., Holm, A., & Breen, R. (August 12, 2010). Comparing Regression Coefficients Between Models using Logit and Probit: A New Method. CSER Working Paper Series, WP No. 0003. Centre for Strategic Educational Research: Aarhus, Denmark: DPU, Aarhus University.http://www.cser.dk/fileadmin/www.cser.dk/wp_003kbkkkjrb.pdf
>
> Mood, C. (2010). Logistic regression: Why we cannot do what we think we can do, and what we can do about it. European Sociological Review, 26(1), 67-82.
>
> Williams, R. (2009). Using heterogeneous choice models to compare logit and probit coefficients across groups. Sociological Methods & Research, 37(4), 531-559.
>
> Williams, R. (September 2011). Comparing Logit and Probit Coefficients between Models and Across Groups. Notre Dame Sociology.http://nd.edu/~rwilliam/stats/Oglm.pdfhttp://www.nd.edu/~rwilliam/stats/Oglm-Handout.pdf
>
> Long, J.S. (2009). Group Comparisons in Logit and Probit Using Predicted Probabilities. Working paper draft 2009-06-25.http://www.indiana.edu/~jslsoc/files_research/groupdif/groupwithprobabilities/groups-with-prob-2009-06-25.pdfhttp://www.indiana.edu/~jslsoc/research_groupdif.htm

Anestis Karvounidis

I would like to pose a question regarding the comparison of the
results of conditional logistic regression. I want to check for the
moderatoring effect of type of business activity (if the company has
manufacture or service activity in host country) in distance factors
(Culture, geographic, economic). Given that the model was non
convergent when I was including the interaction variables, I forced to
run three times the same regression (Conditional logistic regression)
with the same variables but with part of the initial sample. More
specifically, the first model has the total sample, while the rest two
models have the one only the part of sample with manufacture activity
and the other the part with service activities.
>> My point is to check if the type of business activity moderated or intensify distance effect. As a result I want to compare the coefficients of these 3 models. Can I do it directly that, or should I do any additional test. I found suest order of STATA but it does not work, probably because I have different size of sample. Should I prove that the change is statisticaly significant? How can I do it that?

*
*   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/