Dear Joao Santos Silva , Tom Zylkin and Jeff Wooldridge,
I am a master's student from India. After referring to previous posts, I used ivpoisson with the control function option. I have a count variable Y that is non-zero and takes positive values. Both my instrument and endogenous variable are binary variables that are time-invariant. I have panel data with Y over 5 years after the treatment was assigned. In my IV Poisson estimation, I would like to add year and region fixed effects. The total number of observations is 7000, with the unit of analysis being a village, distributed across 300 regions. However, including region dummy variables is computationally intensive due to the large number of regions. Is there a more efficient way to overcome this problem?
As an alternative, can I use this to define a control function?
I am a master's student from India. After referring to previous posts, I used ivpoisson with the control function option. I have a count variable Y that is non-zero and takes positive values. Both my instrument and endogenous variable are binary variables that are time-invariant. I have panel data with Y over 5 years after the treatment was assigned. In my IV Poisson estimation, I would like to add year and region fixed effects. The total number of observations is 7000, with the unit of analysis being a village, distributed across 300 regions. However, including region dummy variables is computationally intensive due to the large number of regions. Is there a more efficient way to overcome this problem?
As an alternative, can I use this to define a control function?
Code:
reghdfe x z $cntrl, a(region year) resid predict vhat, residuals eststo y4 : poisson y x $cntrl vhat i.a_blk_code i.year, vce(bootstrap , reps(1000))
