Hello, I have been computing average treatment effects (ATE) and local average treatment effects (LATE) for an RCT.
For the ATE, I run the following regression 1:
where treatment is equal to 1 for control, 2 for treatment 1, and 3 for treatment 2.
When I tried computing LATE using the following regression 2:
I get the error that the endogenous variables may not be interactions or factor variables
Thus I generate separately the interactions and run regression 3:
My question is there something unsound with the way I go about it in regression #3? Is there a reason why the Stata command does not allow interactions and factor variables for the endogenous variable?
Thank you in advance for the help.
For the ATE, I run the following regression 1:
Code:
xtreg case_clerance i.treatment controls , re
When I tried computing LATE using the following regression 2:
Code:
xtivreg case_clerance controls (i.treatment#c.take-updummy = i.treatment) , re
Thus I generate separately the interactions and run regression 3:
Code:
xtivreg case_clerance controls (treatment1_takeup treatment2_takeup= i.treatment) , re
My question is there something unsound with the way I go about it in regression #3? Is there a reason why the Stata command does not allow interactions and factor variables for the endogenous variable?
Thank you in advance for the help.
