-ppml- now available in SSC

[Joao Santos Silva]

Dear All,

Thanks to Kit Baum, a new version of -ppml- is now available in SSC. This version is only slightly different form the original version published almost 5 years ago in the Stata Journal, but users are encouraged to update. Please do let me know if you have problems using the new version.

Thank you for your interest in -ppml-,

Joao

[Neha Kanojia]

Dear Joao Santos Silva
I am estimating a bilateral gravity model where source countries are 10 in numbers and the host country is only one. I have many zeros in my dependent variable. The time period is 15 years so my dataset is small so can I apply ppml or ppmlhdfe model? I am applying the augmented gravity model, is it fine or should go for any other model?

I am asking this question as I have read that ppml can be applied in the big dataset with heteroscedasticity.
please provide some references to study about the ppml and ppmlhdfe model in details.

[Joao Santos Silva]

Dear Neha Kanojia,

Your dataset is really small, but I would still use the PPML estimator, as implemented in ppmlhdfe. However, I would consider expanding your dataset because otherwise the results may not be reliable.

Best wishes,

Joao

[Neha Kanojia]

Dear Joao Santos Silva

Thanks a lot for your response.
The host country in the study is India and the source countries are ASEAN countries. As I am studying a block, I cannot increase the dataset.
I have two doubts
1) If I apply the ppmlhdfe , can I do something to make my estimators reliable?
2) As per my little knowledge, I think ppml command provides inbuilt robust results. Is ppmlhdfe also give an inbuilt robust result or do we have to use a command to get robust results?

[Joao Santos Silva]

Dear Neha Kanojia,

The results of the two commands should be identical, but ppmlhdfe is faster and is better at dealing with some numerical issues.

Best wishes,

Joao

[Neha Kanojia]

Dear Joao Santos Silva,

Thanks a lot for your prompt response.

[Neha Kanojia]

Dear Joao Santos Silva,

Can you plz tell me the commands which we have to follow while applying ppmlhdfe ?, I am using this model for the first time. I need your guidance.

[Joao Santos Silva]

Please check the help file for the ppmlhdfe command for examples.

Best wishes,

Joao

[Neha Kanojia]

Dear Joao Santos Silva

the command which I am getting in the help file is as follows which is giving robust results directly-
CASE 1-
ppmlhdfe depvar indepvar, absorb( panel id)

but I am confused should I apply the above command or
CASE 2-

apply the Hausman test first

xtpossion depvar indevar, fe
estimate store
xtpossion depvar indepvar, re
hausman fe re, sigmamore

then on the basis of hausman test apply fixed or random effect robust results.

xtpossion depvar inpvar, fe vce(robust)

please guide which one is correct?

[Joao Santos Silva]

There is no need to use the Hausman test, just use the estimator with fixed effects. In the contxt of non-linear models such as the one estimated by PPML, the random effect estimators are not interesting because they depend on very strong assumptions.

[Neha Kanojia]

Dear Joao Santos Silva
Thanks a lot, I will run the command then I will show you the results.

[Neha Kanojia]

Dear Joao Santos Silva

I have attached the estimated results for ppmlhdfe, can you please check and suggest are they correct?
I have checked multicollinearity and directly apply the ppmlhdfe command, can we check the significance level of other time-invariant variables by applying any command?

[Joao Santos Silva]

I do not know what you are doing, so I cannot make any detailed comments, but I do not see anything suspicious.

[Neha Kanojia]

Thanks a lot for your guidance. It is really helpful.

[Neha Kanojia]

Dear Joao Santos Silva,

As discussed with you, I applied ppmlhdfe model to estimate a bilateral gravity model where source countries are 10 in numbers and the host country is only one. I have many zeros in my dependent variable and the time period is 15 years. I have 2 questions:-

1)Do I need to apply any test to check the specification of the model like Ramsey reset Test and Manning and Mullahy test before application of ppmlhdfe?
2) command applied by me is correct or not?

For getting, autocorrelation corrected results

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

OR

for getting, autocorrelation and heteroscedasticity corrected results

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

please guide.