Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: problematic mlogit results


From   Uygar Yuzereroglu <uygach@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: problematic mlogit results
Date   Mon, 11 Apr 2005 12:13:07 -0700 (PDT)

Hi all,

I have a kind of specific question. I am working on a
multinomial logit model with social interactions, in
line with the work of Durlauf. (I already grouped the
people in my dataset according to their socioeconomic
status.) However the estimation results are kind of
strange. 

To see the social interactions (endogeneous and
contextual effects) I have four variables: 
- The percentage of people making the same choice as
oneself within his reference group 
- Three variables of group averages (average culture,
education and income levels within the reference group
of oneself)

In the results, the coefficients of all choices except
one seems ok. For one of the choices the vairance of
all estimated coefficients are so big that the
associated p-values are equal to 1. 

The reason that I considered was a high correlation
between these variables within the group of people
having this problematic choice. However, it is the
contrary: the correlations within the groups of people
with other choices are much higher, in fact around
0.9. On the other hand, within this problematic choice
the correlations are all less than 0.5.

Is there anyone who can figure out what the problem
might be? Can it be multicollinearity? If needed, I
can send my estimation output, and discuss more about
the subject, which I need desperately :)

Thanks.
Uygar


		
__________________________________ 
Do you Yahoo!? 
Yahoo! Mail - Helps protect you from nasty viruses. 
http://promotions.yahoo.com/new_mail
*
*   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