Panel data analysis for specific countries in a dataset

[Shayan Mujtaba]

Hi, I have a question regarding doing a regression analysis on a panel dataset. I have gathered data on 38 countries in my dataset with multiple variables for 11 years from 2010 to 2019. I am able to do a fixed vs random effects estimation on my model, but I was curious if it is possible for me to do the analysis on selected countries of my choice from the dataset. For example out of the 38 countries I have, if I wanted to do the analysis on just Japan, Korea and China from my model, is that possible to do that in STATA using the same dataset? I would like to be able to interpret all my coefficients in my model for each country (is what I am trying to ask). Is that possible? Any help would be greatly appreciated.

[Carlo Lazzaro]

Shayan:
yes, you can do it using an -if- qualifier aimed at selecting the three countries you're interested in.
See -help if- (qualifier).

[Shayan Mujtaba]

Hi Carlo, I did use the if qualifier but STATA simply omits some of my variables because of collinearity and I get no values for the standard errors, t statistics and p values. I really appreciate you response.

[Carlo Lazzaro]

Shayan:
are you using the -fe- estimator?

[Shayan Mujtaba]

Yes I am using the -fe- estimator. I have already done the hausman test using both -fe- and -re- estimators and the test suggests that the -fe- estimation is more appropriate for my model. The only thing I would like to do more is to be able to do a country wise analysis. Because I am curious to find out how the individual countries are effected by my chosen model.

[Carlo Lazzaro]

Shayan:
1) the -fe- estimator wipes out time-invariant variables;
2) to work as its best, the -fe- estimator needs a relevant within-panel variation in the time-varying variables;
3) if you're interested in how the individual countries (that is, taking one country at a time) are effected by your chosen model, you should run an OLS.

[Shayan Mujtaba]

Ah, thank you so much for the clarification Carlo. I will do that.
Regards,
Shayan