[Joao Santos Silva]

Dear Neha Kanojia,

1) You can perform the RESET test after estimating the model; I would not use any other tests.
2) The commands look good; you should cluster, but 10 countries is not enough for clustering. You should use a larger data set.

Best wishes,

Joao

[Neha Kanojia]

Dear Joao Santos Silva,

My study is related to block i.e. the ASEAN(Association of Southeast Asian countries ) therefore I am taking 10 countries only. please suggest should I use vce clustering or vce robust. As in the log of gravity paper, I have studied that results in the ppml are heteroscedasticity corrected.

Best Regards,
Neha

[Joao Santos Silva]

Can you use 10 host countries rather than just 1? That would give you enough clusters.

[Neha Kanojia]

I am studying the determinants of FDI inflows from ASEAN in India. due to this Objective, I have only one host country. I can add one more Objective i.e. I can study the determinants of intra-FDI inflows in ASEAN, here host countries can be 10.

I have a doubt that can we use panelid with absorb with one host country in the below stated stata command and you are talking about the clustering in vce command right?

ppmlhdfe depvar indepvar, absorb( panelid) vce(robust)

Best Regards
Neha

[Joao Santos Silva]

If you use 10 hosts, you should absorb the so called origin and destination fixed effects, not the panelid; you should cluster by panel id.

Can you please say more about the purpose of the study? Is this for a masters dissertation?

Best wishes,

Joao

[Neha Kanojia]

Dear Joao Santos Silva
This is one of the empirical chapters of my Ph.D., this chapter of the study tries to examine the important determinants of the FDI inflows in India from ASEAN countries for the period 2004-2019. India is a destination country and ASEAN countries are source countries. Most of the variables in the study are taken in the form of ratios( home country variable/host country variable)

Panel id is the numerical value that I have given to the source countries. As I have taken variables in the ratios therefore I have not given any numerical value to the destination country.

ALL THE SUGGESTIONS AND GUIDANCE GIVEN BY YOU ARE VERY HELPFUL, I WILL TRY TO INCORPORATE THEM. I WILL SHOW YOU THE FINAL RESULTS. THANKS, A LOT.

Best Regards
Neha

[Neha Kanojia]

I have tried Ramsey reset test after ppmlhdfe by applying
estat ovtest command but it's showing error.
if I am applying estat ovtest after pooled ols then it's working.
Please guide. What is the correct way and command for applying Ramsey test for ppmlhdfe.
Best regards
Neha

[Joao Santos Silva]

The ovtest command is not a good way of performing the RESET test, and it only works after OLS. You can find it how to do the test here.

[Neha Kanojia]

Thanks for the Link of Ramsey reset test,
after application of the test on the robust model of ppmlhdfe,
I got these results

Chi2( 1) = 1.46
prob>chi2= 0.2269

is this means that the model is not fit to apply ppmlhdfe ?

[Neha Kanojia]

What would be the null hypothesis for Ramsey reset test ?

[Joao Santos Silva]

The null is that the model is correctly specified.

[Neha Kanojia]

Dear Joao Santos Silva,

I have estimated the results in 2 cases and attached the same, please check the estimated results.

CASE 1 where I have dropped the common language and common colony from my dataset
CASE 2 where I have taken common language and common colony in the dataset

for estimating the results, I follow the following commands:
xtset panelid Time

correlate indepvar
regress depvar indepvar , vce(robust)
estat vif

xtserial indepvar

xtreg depvar indepvar, fe
xttest3

ppmlhdfe depvar indepvar , absorb( panelid) vce( robust)
predict fit,xb
gen fit2=fit^2
ppmlhdfe depvar indepvar fit2, absorb( panelid) vce( robust)
test fit2=0

please check and give your feedback for
1) whether the commands used are correct and sufficient to estimate the results?
2) results estimated in the attached files are correct or not?

[Joao Santos Silva]

Dear Neha Kanojia,

I do not really know what you are doing, so I cannot comment in detail. You should discuss the results with your supervisor.

Best wishes,

Joao

[Neha Kanojia]

can you please comment about the commands used in the model?

Best Regards
Neha

[Joao Santos Silva]

ppmlhdfe is the right command to use.