in my panel T=8 and N= 108, which model will be preferred: xtgls, xtpcse or xtreg
More information than the numbers you provide are needed to answer this question. Assuming this is stata you are discussing, I would recommend starting with the basic xtreg.
In this case the dynamic panel and the GMM method are more appropiate. The command with stata is xtabond2
Dr. Dahora:
Your question is quite similar to that posted by Florian at the Statalist archive: http://www.stata.com/statalist/archive/2010-01/msg00430.html.
Unfortunately, it came without a proper answer: http://www.stata.com/statalist/archive/2010-01/msg00431.html.
As pointed-out there, we also need more information to help you better. However, you will find some interesting insights here: http://www.stata.com/statalist/archive/2004-04/msg00729.html, in order to choose between xtgls and xtpce. I think that xtreg (fe, re, or be) will be appropiate for your small database (because iid residuals usually appears with large T and N).
Finally, I strongly recommed to read this paper:
http://jjxy.znufe.edu.cn/xxgg/201110/P020111014344298906301.pdf (pages 998-999)
Best regards
Excuse-me Madan
There was a typos mistake. Please, replace:
"I think that xtreg (fe, re, or be) will be appropiate for your small database (because iid residuals usually appears with large T and N).".... by
I think that xtreg (fe, re, or be) will NOT be appropiate for your small database (because iid residuals usually appears with large T and N).
sorry
With this short information, I think that xtreg is preferred, but if it's possible you could enlarge your time series, because you need more time period to observe the possible inconsistencies.
Hi. If you are using country level data then because you have more N in comparison to T, using xtgls and xtpcse will not produce the expected results. xtgls and xtpcse are used with long panel, where T is not much larger than N. In your case, you have larger N than T, so you have short panel. Therefore, I would suggest you use models suggested for short panels, such as xtreg, re vce(cluster id). For more information I would suggest you read "Microeconometrics Using Stata" by Pravin K. Trivedi · A. Colin Cameron. I am sure it will provide a great solution.
Ikboljon Kasimov please I have the same question , I have random panel data and N>T i tried both xtpcse (we use this command when the N>T ) and the cluster option and i find two different result. please could you advise me which is the correct method i have to use . Thank you :)
@ikboljon kasimov my N=119 and T=16 years. I found the problem of hetroscedasticity, autocorrelation and cross sectional dependency. Is it appropriate to apply xtpsce.
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India