[Eric de Souza]

"I assume that we should set the panel at the highest aggregate level, here, province"

If you have a certain overlap between industries and provinces: certain industries are present in certain provinces; then there is no hierarchy between industries and provinces
I did not take this into account in my previous reply (#15)
On Edit: which is therefore incomplete and I would have to think about it

[lal mohan kumar]

Thanks Eric de Souza
You mean

 
 xtset province year   // I assume that we should set the panel at the highest aggregate level, here, province egen province_year=group(province year)       // grouping province and year  xtreg depvar indep var i.industry i.province_year i.year,fe

Am I correct now?

[Eric de Souza]

I need to think about it

[lal mohan kumar]

Thanks Eric de Souza. In case you find something interesting, kindly post it

[Musa Isa]

I have a somewhat similar concern. I am working on the effect of rainfall anomalies on conflict in Nigeria with 37 cohorts (states) from 1997-2021. I run reg y x i. year the results shows a significant relationship. However, when I include i.state in the model, the relationship disppear. I switch to xteg,fe the result still shows insignificant relationship. My question is, can I proceed to with the re y x i.year model?

[Musa Isa]

I have a somewhat similar concern. I am working on the effect of rainfall anomalies on conflict in Nigeria with 37 cohorts (states) from 1997-2021. I run reg y x i. year the results shows a significant relationship. However, when I include i.state in the model, the relationship disppear. I switch to xteg,fe the result still shows insignificant relationship. My question is, can I proceed to with the re y x i.year model?

[Carlo Lazzaro]

Musa:
you're seemingly hunting for the "best" model that can fullfil your expectations.
This fishng expeditions are not the way to go.
You are dealing with a N>T panel dataset.
Therefore, if your regressand is continuous, you should go -xtreg,fe- first.
If you have other issues, please get back to this forum sharing what you tyoed and what Stata gave you back (as per FAQ). Thanks.

[Musa Isa]

Hi Carlo, I am dealing with a binary regressand, I realized that I mistakenly used the wrong data, I fixed it and it worked. Thank you for the response.

[Carlo Lazzaro]

Musa:
thanks for explaining what went wrong at your first attempt.