[Clyde Schechter]

Yes, the modification to restrict to RD_ratio > 0 you propose is correct.

If i use p33 and p66 to group my banks, then for my regression i have to run two separate regressions, one with if RD_ratio_group = 2 and one with if RD_ratio_group = 1?

At this point I don't clearly recall (if you ever even explained it) exactly what kind of analyses you are planning and what role these RD ratio groups play in it. Separate regressions is one possibility, but depending on how you plan to use the results, it may be better to run a single regression in which everything (or all the important predictors) are interacted with i.RD_ratio_group). In particular, if you plan to specifically contrast regression parameters across the RD ratio groups, the interaction approach will make that easier.

[Vasilis Papas]

Previously, when i was using the median to create my bank groups, my regression was

xtreg Loan_ratio Loan_ratio_lagged c.EL_ratio_lagged##NIRP##c.RD_ratio_lagged L1.Dep_Riks_ratio L1.Debt_sec_ratio i.bankid i.mdate if RD_ratio_group== 0 & mdate< tm(2018m12), fe  vce(cluster bankid)

and

xtreg Loan_ratio Loan_ratio_lagged c.EL_ratio_lagged##NIRP##c.RD_ratio_lagged L1.Dep_Riks_ratio L1.Debt_sec_ratio i.bankid i.mdate if RD_ratio_group== 1 & mdate< tm(2018m12), fe  vce(cluster bankid)

and i was comparing the effect of NIRP on loan ratio for high and low RD_ratio banks. So if i understood correctly, if i separate my banks with respect to p33 and p66, now i can just add a third regression with for the last group of RD_ratio_group = 2 or i can write the regression as

xtreg Loan_ratio Loan_ratio_lagged c.EL_ratio_lagged##NIRP##c.RD_ratio_lagged#RD_ratio_group L1.Dep_Riks_ratio L1.Debt_sec_ratio i.bankid i.mdate if high_RD_ratio_all_months2 == 0 & mdate< tm(2018m12), fe  vce(cluster bankid)

[Clyde Schechter]

If the approach you used with high and low RD_ratio banks (split at the median) was satisfactory, then your proposed approach will be, too.

[Vasilis Papas]

Dear Clyde Schechter thank you again so much for all your help and your time. Your help has been extremely valuable!