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

Re: st: Model choice for predicting ordered non-normal categorical variable


From   John Hendrickx <john_hendrickx@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Model choice for predicting ordered non-normal categorical variable
Date   Wed, 30 Oct 2002 07:35:49 -0800 (PST)

You might want to try a Stereotyped Ordered Regression (SOR) model. A
SOR model calculates scale values for the dependent variable and
effects of the independent variable on the scaled dependent variable.
I'm not sure it would work with 18 categories though unless you have
a lot of data (and a fast computer).

A SOR model can be estimated using the -mcl- package, available from
SSC. Information is available at
http://www.sls.wau.nl/bk/bedrijfskunde/jhendrickx/mcl/stata/ in STB
56, and in briefly in Long (2001), "Regression models for categorical
dependent variables using Stata", paragraph 5.9.2.

Hope this helps,
John Hendrickx

--- Scott Winship <swinship@wjh.harvard.edu> wrote:
> I am trying to estimate a model predicting "food security" (a
> construct
> ranging roughly from adequate food levels with no insecurity to
> severe
> hunger).  Food security is a latent variable, where I observe
> whether or not
> a household experienced each of 18 food-related problems.  The
> problems
> generally are of increasing severity (decreasing prevalence).  A
> sizable
> majority of households experience none of the 18 problems.
> 
> I have so far estimated OLS and ordered probit models, which I
> realize are
> not ideal.  I have also estimated a probit model predicting whether
> a
> household experiences 0 or more than 0 problems along with an
> ordered probit
> model for those households experiencing at least one problem.  Have
> been
> told to consider a negative binomial or zero-inflated poisson
> regression
> model, but these don't seem quite right (given that I'm not looking
> at
> counts of independent events).  Seems like I'd want a model similar
> to an
> ordered probit but with an assumed latent distribution that was
> non-normal.
> 
> Any suggestions are appreciated.  - Scott
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>      Scott Winship
>      Ph.D. Candidate in
>      Sociology & Social Policy
>      Harvard University
>      swinship@wjh.harvard.edu
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> *
> *   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/


__________________________________________________
Do you Yahoo!?
HotJobs - Search new jobs daily now
http://hotjobs.yahoo.com/
*
*   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