Hi,
I have a question regarding panel data with T > N. I am using the panel regression method, and I read in some journals that I need to apply the cross-sectional dependence test.
I found in a Stata journal that if the panel data is T>N (17>10), I have to use the LM Breusch Pagan test with Stata command xttest2. If there is cross-sectional dependence in my data, how do I correct it? Is there any way to fix it or do I have to change the method?
Thanks
Hello Ruth Mahya, I suggest you first run other cross-sectional dependency (CSD) tests to confirm the results of the Breusch Pagan LM test. Also, you don't necessarily have to change your model or correct for CSD, you rather need to use estimation techniques that account for CSD. A good example will be Driscoll-Kraay Standard Errors (DKSE) or Panel-Corrected Standard Errors (PCSE). My humble suggestion.
Ruth Mahya, since you found CSD, add methods (DSKE) as suggested by Gabriel Mordzifa Sackitey to strengthen your model's robustness. Cheers.
Use tests like Pesaran CD, Breusch-Pagan LM, or Friedman test. If T>N, CD tests are more reliable.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism