Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Choosing estimators for xtnbreg (re, fe, pa?)

From   yuchian ng <>
Subject   st: Choosing estimators for xtnbreg (re, fe, pa?)
Date   Wed, 27 Jul 2011 20:22:41 +0800

Dear Statalisters,

I am quite confused by the different estimators and such that I can
choose from when using the negative binomial regression, for handling
panel data. Some examples that I have come across are:

1) xtnbreg, re (I know this is the default one)
2) xtnbreg, fe (i know this is the fixed effects estimator, but I am
still confused after reading quite a bit of material)
3) xtnbreg, pa (i know this is called a "population-averaged" model,
but I have no idea how this works!)
4) nbreg, fe cluster(country)

Specifically, I am hoping to generate a regression that has:

Dependent Variable: Number of Terrorist Incidents in a year
Independent Variables: Democracy (whether the country is a democracy
or not), Import levels, Export levels, dummy variables for each year

My questions are:

a) May I ask if there is any concise explanation for how I should
decide which model would best suit me, or if anyone knows of a good
resource that I can refer to?

b) Also, What are the differences between the "random", "fixed"
effects models and the "population-averaged" model?

c) Is there a difference in using "xtnbreg" and "nbreg, cluster(country)"?

d) Why am I unable to use "estat ic" after using "xtnbreg, pa"?

e) Why am I unable to use "vce(robust)" for the "xtnbreg" models?

f) Is it necessary to put dummy variables for each year of the data in
my panel, in order to control for time-variable effects, when using
the "fe" model?

Thank you so much for all your help!! I am desperately in need of guidance!

Yours Sincerely,
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index