Hello,

I have panel data for a sample of European banks. However, the sample is not really representative right now as one country, Italy, has more than half of all observations. Therefore, I wanted to exclude 5% of the smallest Italian banks in order to make the sample more representative of the population of interest, that is, the EU banking sector. Right now, I have a dummy=1 if a bank is Italian, but I cannot figure out how to remove 5% of the smallest observations (by bank size) only for Italian banks. Additionally, I was thinking about whether I should remove the observations by year or first consolidate bank assets over all the years and then remove the 5% of Italian banks that had the smallest amount of assets over the time period?
Currently I have a balanced panel with 525 banks, 267 of these are Italian and I have 9-time observations for each bank.

I'm really stuck on this and I cannot seem to find any code that would be applicable in my case, so I would greatly appreciate your input!