Regression specification

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Regression specification

Florian Esser
This post has NOT been accepted by the mailing list yet.
Hi everyone,
being quite new to (advanced) econometrics I am struggeling a bit about the seemingly unlimited STATA commands to chose from when estimating my model.
I have a strongly balanced panel with large N (101) and small T (6 years).
In order to determine the "correct" estimation model, I run several tests:
1. Heteroscedasticity: xttest3 ---> Heteroscedasticity present
2. Autocorrelation: xtserial ---> Autocorr present
3. RE vs. FE: xtoverid..., robust ---> P-vaue = 0.000 (do I understand it correctly, that this means using FE?)
4. Cross-sectional dependence: xtcsd, pesaran abs and xtcsd, frees ---> CSD present

Based on this, I would use xtscc..., fe to use Driscoll Kraay standard errors.

I have several additional questions:
1. If I were to include the lagged independent variable as a dependent variable would I have to use xtabond2? What would be the alternatives?
2. When can I use xtgee? Is this just appropriate for RE-models?
3. Is it correct to use xtpcse only for long panels? Why would I use it instead of, e.g. xtgls?
4. I have rarely seen paper (in strategy research) testing for time-fixed effects (testparm for the Year dummies). How "required" is this?
5. Are there any other tests I should run?

I know that are a lot of - and probably beginner - questions - but your help is greatly appreciated.
Thanks in advance!
Reply | Threaded
Open this post in threaded view
|

Re: Regression specification

Florian Esser
This post has NOT been accepted by the mailing list yet.
Is there an edit function?

In question 1, I involuntarily exchanged the independent and dependent variable.
It should be:
"1. If I were to include the lagged dependent  variable as an independent variable would I have to use xtabond2? What would be the alternatives?