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

Re: st: Countvariables/Poisson Regression in hetergenous choice models

From   Maarten buis <>
Subject   Re: st: Countvariables/Poisson Regression in hetergenous choice models
Date   Tue, 24 Jun 2008 22:22:06 +0100 (BST)

--- Sebastian Sattler <> wrote:
> I am searching a procedure which helps me to estimate the networksize
> of respondents (count variable).
> I want to compare the coefficients of three different studies.
> Williams provided a solution for binary and ordinal dependent
> variables with his oglm-procedure (heterogeneous choice model). OLS
> regressors can be compared easyly - heteroscedasticity is not that
> problem.
> But I do not find any for count outcomes.
> Another connected problem in the same context is the "scale" of my
> dependent variables. In one study the networksize is restricted to a
> maximum of 3 (range 0-3). The other studies do not have any
> restriction. There empirical range is: 0-20 and 0 to 11 (some
> networkgenerators produces a range from 0 to 8).

The default initial choice for a model of a count variable is Poisson
regression, available within Stata as -poisson-. Your reference to
-oglm- suggests that you have some additional variance that you want to
model in addition to modeling the networksize. That can be achieved
through -nbreg-. Your second question is harder to solve, in principle
the problem should be solvable, in particular by using the information
from the different studies together should help in the identification
of such a model, but you'll probably have to program the likelihood
yourself using -ml-. If you decide to go on that road than you can't
live without:

Hope this helps,

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Sent from Yahoo! Mail.
A Smarter Email
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index