[Meng Zhang]

Dear Dias,
Thank you for your reply.
I did not say it clearly in the last post. I also want to use PPML with fixed effects or random effects. The code in the last seems only pooled regression.

egen id=group(countrycode sic2digit)
xtset id year

fixed-effects:

xi, pre(id_) noomit i.id
xi, pre(y_) noomit i.id

ppml ofdi lndist contig lngdp lnchngdp countryskill industryskill countryskillindustryskill countryrdpeople industry_rdintensity countryrdindustryrd bits fta cu economicfreedomindex financedevelopment id_* y_*, cluster(id)

Is this code right? In this code, id is a combination of country and industry, so once I included "id_*" in the code, I do not have to include country_* or industry_*, because they had been fixed through "id_*", right?

Then how to write code with RE?

All the Best,
Meng

[Said Jafar]

Dear Meng,

Note that you have xtset your data, so it is panel regression, not pooled
I am not really sure, but it seems your code is correct.
Yes, you should not include country_* or industry_*, they will be omitted.
I would say, no one uses random, so it is better stick on fixed.

Best