Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# st: Logit: (un)conditional fixed effect and clustering

 From 德维园北京 To statalist@hsphsun2.harvard.edu Subject st: Logit: (un)conditional fixed effect and clustering Date Thu, 7 Mar 2013 17:24:04 +0100

```Dear Statalisters,

I have a pseudo-panel data with observations of activities of firms
over years: about 570 firms, over 10 years and activities within a
firm vary in numbers (all together I have more than 60,000
observations of activities, unequally distributed across firms).

The dependent variable is dichotomous, so I run simple logit
regression with robust standard errors clustered at the firm level.
Year dummies are included. The results are fine.

I read that I could also run fixed effect logit: both conditional and
unconditional. I am a bit puzzled what they are and how to do them.

My syntax:
- For simple logit
logit y x1 x2 x1x2 ..., robust cluster (firmid)

- For conditional fixed effect:
clogit y x1 x2 x1x2 ..., group(firmid)

- For unconditional fixed effect:
logit y x1 x2 x1x2 ... i.firmid, robust cluster (firmid)

In the second and third models, observations without within-group
variance are dropped.

Is this the right way to process? I heard that the simple logit with
robust standard error might be a comprise because of complications of
both fixed effect models. I appreciate if someone can explain the
difference between "conditional", "unconditional" fixed effect and
simple logit with clustered standard error. Thank you!

Best,
Chris
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
```