Hello, I am dealing with a econometric model that uses system GMM. Please answer the following questions: 1.In order to get better result, what tests should I accomplish before running the final regression? I mean do I need to test autocorrelation, heteroscedasticity etc. for diffferent series of the panel data ? 2.What is the appropriate test to determine the endogeneous,exogeneous and predetermined variable in the panel data? How these issues can be resolved?
You don't test for endogenity or exogenity. You must have a priori knowledge of the endogenous as well as the exogenous variables. The response variable is the endogenous one and the explanatory variables the exogenous ones. There is no test to identify them. The are known a priori.
For system GMM, the convention is usually to undertake post-estimation diagnostic tests. The following tests apply: (a) A test for zero autocorrelation in first differenced errors (e.g. Arellano Bond test). In this case Ho=there is no autocorrelation. (b) A test for over-identifying restrictions (eg Sargan test). In this case Ho=over identifying restrictions are valid. (c) A test for joint significance of parameters (e.g. Wald Chi sq test)
You have to test the post-estimation diagnostic test, which includes; Then test for zero autocorrelation in first differenced, the Sargan test for over-identification restriction and the Wald Chi Sq test for joint significance of parameter.
Check stata forum for procedure and how to apply stata for estimations and test;
https://www.statalist.org/forums/forum/...stata.../1401573-xtabond2-and-deeper-lags
**
Goodluck
Please can some one tell me the difference between GMM ans System GMM, because to my knowledge I believe that System GMM is sub-test under GMM. Thanks
dear Adeley E. Oloniluyi
for your question, please refer to this
https://www.stata-journal.com/article.html?article=st0159
regards,
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running