Subject: st: Is it necessary to have some variable for exposure in xtnbreg

[jhilbe@aol.com]

I get the digest, which is a day later, so someone may have already 
answered this query, in which case I
apologize for the repetition.

You do not need to have an exposure variable for xtnbreg. You do need, 
however, a variable that identifies
the panel structure of the model, whether it be a GEE, random effects, 
or fixed effects model.  Standard Poisson
and negative binomial models do not have to have an exposure either, 
which is simply another word for
an offset where the offset variable is logged.

Joseph Hilbe






Date: Tue, 14 Sep 2010 20:33:33 -0400
From: Indu Khurana <indu.statauser@gmail.com>
Subject: st: Is it necessary to have some variable for exposure in 
xtnbreg

Dear Statalisters,

I have a count data on number of patents which is the dependent
variable (panel setting). Right hand side variables include Employee
well being and debt-equity ratio. The data on patents has a lot of
zeros. Although ZINB could have been an option but I am using fixed
effects. Therefore I decided to settle with negative binomial, fixed
effects - xtnbreg

My question is: Is it necessary to have some variable for exposure in 
xtnbreg?

The data is on 5000 firms for 5 years which all came into existence at
the same time in 2004. So basically they all have the same exposure.
And there is no information on month of the the year.
There is no other time dimension which can be used to fit in exposure.

I was thinking of the following options:
1. To ignore exposure variable all together
    Which I am not sure of whether it is correct.
2. To include number of years of owner's education to fill as exposure.



Please suggest.

Thanks

Indu Khurana



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/