Regressions one bank per country over 15 years

[Moh Bit]

Hello, I have have a sample of banks in 68 countries over 15 years. The sample is dominated by US and German banks. In my robustness tests, I am trying to run the regressions with one bank per country over 15 years. Can you please advise on how I can run such a regression?

[Clyde Schechter]

Assuming you want to select one bank at random from each country and use the same bank for all 15 years:

set seed 1234 // OR ANY OTHER RANDOM NUMBER SEED

gen double shuffle = runiform()
by country (shuffle), sort: gen byte to_use = _n == 1
by country (to_use), sort: keep if bank == bank[_N]
// NOW YOU CAN RUN YOUR REGRESSION

Added: Actually, the above can be shortened to:

set seed 1234
gen double shuffle = runiform()
by country (shuffle), sort: keep if bank == bank[1]

[Moh Bit]

Thank you, Clyde!