Re: st: Categorical dependent variables and large dummy variable datasets

[David Jacobs <jacobs.184@osu.edu>]

Perhaps I'm missing something important in your question, but doesn't the 
cluster option for the estimation procedures you wish to use suffice?  For 
example, after an ologit or mlogit equation, simply put a clause ", 
cluster(cluster ID #)" with the parenthetic expression containing the 
variable name for the id number of your clusters.  By using the cluster 
option in Stata, you automatically get Huber/White robust standard 
errors.  You might also look at the survey estimators in the manuals.

Dave Jacobs

At 10:22 AM 12/18/2002 -0500, you wrote:

> I work with cluster sampled survey data and estimate models with mainly
> ordinal and nominal level dependent variables. I understand that the
> standard errors in these models are affected by the fact that the
> observations are not truly independent (i.e. grouped by clusters). I
> cannot account for this simply by adding dummy variables in the model
> because I have approximately 400 clusters.
>
> What is the best way to handle this? If I were using OLS to estimate the
> models; areg, absorb(cluster) would seem to be the way to go. However, I
> need to use variants of logistic regression (mlogit, ologit).
>
> I have also looked into using the ,robust option to calculate robust
> standard errors. Will this provide correct standard error estimates?
>
> Thank you.
>
>
>
> Tom Brewer
> Department of Justice Studies
> Kent State University
> 113 Bowman Hall
> Kent, OH 44242
> 330-672-0315
>
> *
> *   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/

*
*   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/