Dear all,
I have a question regarding modelling treatment effect heterogeneity on the user-written command rdrobust.
Suppose the running variable is distance to the border. Distance in control states is negative, so that the border is distance = 0. The dependent variable is y.
For now I have:
I have reason to believe that the treatment effect is stronger in states with higher density. So I generate a dummy variable called
.
Therefore, I would like to interact the running variable, distance, with this dummy to model this treatment effect heterogeneity. I have two questions:
- How can I get a coefficient on
and
through rdrobust to model the heterogenous treatment effect?
- Is there anyway to "manually" run a regression discontinuity estimation (simply a first order polynomial) modelling this heterogeneity in the treatment effect between high and low density states, within an optimal MSE-minimising bandwidth?
Many thanks in advance for your help!
I have a question regarding modelling treatment effect heterogeneity on the user-written command rdrobust.
Suppose the running variable is distance to the border. Distance in control states is negative, so that the border is distance = 0. The dependent variable is y.
For now I have:
Code:
rdrobust y distance, covs(x1 x2 statepairFE*) //the covariates are quite important as I wish to compare only contiguous states
Code:
above_median_density_state
Therefore, I would like to interact the running variable, distance, with this dummy to model this treatment effect heterogeneity. I have two questions:
- How can I get a coefficient on
Code:
distance
Code:
c.distance#1.above_median_density_state
- Is there anyway to "manually" run a regression discontinuity estimation (simply a first order polynomial) modelling this heterogeneity in the treatment effect between high and low density states, within an optimal MSE-minimising bandwidth?
Many thanks in advance for your help!
