ppml_panel_sg: RESET test (p-value)

[Daniel Gomez]

Hi,

I am trying the RESET test (p-value) for a gravity model with fixed effects: (exporter-year, importer-year, pair) as following:
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
ppml_panel_sg TRADE RTA if exporter != importer, ex(exporter) im(importer) y(year) cluster(pairid)
estimates store PPMLfesPAIR

***RESET test

predict fit, xb
generate fit2 = fit^2
ppml_panel_sg TRADE RTA fit2 if exporter != importer, ex(exporter) im(importer) y(year) cluster(pairid)
test fit2 = 0
drop fit*
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

However, I received this mistake:

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
note: fit2 omitted because of collinearity over lhs>0 (creates possible existence issue)
Iterating...





test fit2 = 0
fit2 not found
r(111);

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

What should I do?

Thanks!

[Joao Santos Silva]

If I understand it correctly, your only regressor is the RTA dummy. In this case the RESET as you are doing it does not make sense because the square of the dummy is the dummy itself.

Best wishes,

Joao

[Daniel Gomez]

Thank you Joao!

Yes, that is what is happening. So It means that I cannot apply the RESET test for this case?

[Joao Santos Silva]

You can perform a RESET test if you include the fixed effects on the predictions. That is OK as long as you have enough observations for the estimated fixed effects to be reliable.

[Daniel Gomez]

Thank you Joao!

I think I cannot include the fixed effects on the predictions...

Given the high number of countries in my database, I could not create the pair fixed.
For this reason, I am using the ppml_panel_sg estimator:


ppml_panel_sg TRADE RTA if exporter != importer, ex(exporter) im(importer) y(year) cluster(pairid)