How to test Pooled data and do heteroskedasticity test for random effects?

[shahla ebrahimi]

I would appreciate if you could let me know which Stata's command should be used to estimate a pooled data. In fact,

Chow test rejects fixed effects

Breusch pagan rejects random effects


Therefore, pooled estimation is needed but I don't know what is the difference between these two commands:

regress y x

xtreg y x, pa (population averaged)

Besides, could you please let me know how to test heteroskedasticity for random effects (xxtest3 just tests heteroskedasticity for fixed effects)? I also found that LR test is used to do it but it was for xtgls not for xtreg.

Note: the number of companies are greater than the periods i.e., N>T.
Thanks in advance.

[Carlo Lazzaro]

Shahla:
some comments on your query:
- your first code (-regress-) should show clustered standard errors if you have a panel dataset and a pooled OLS is recommended;
- you should share with the list the outcome of your tests and regression models, along with what you typed (as per FAQ);
- https://economics.stackexchange.com/...vs-time-series will give you some hinnt to test for heteroskedasticity in -xtreg,re-.

[shahla ebrahimi]

Hi Carlo
Thanks a lot. If the code changes into the following, what would be the difference between them? Which one to choose to estimate a pooled data?

regress y x  i.time, vce(cluster id)

xtreg y x, pa (population averaged)

[Carlo Lazzaro]

Shabla:
in general, the approach for estimating a pooled OLS is your first code.