Dear All,
I am doing a causal treatment effect analysis where my treatment variable is multivalued (it has 3 categories) and I am using the IPWRA framework for estimating ATT (teffects ipwra). To ensure robustness of my estimates while using this framework, I have to test that the conditional independence assumption (CIA) and overlapping assumption do hold. How can this be done based on the approach I am using for causal analysis?
2. With the same framework how can I justify that non-observable variables do not have much effect on my estimates? my reason for using IPWRA in a multivalued framework is because I did not have a valid instrument to control for selection on unobservables (Hence it is why I did not use multinomial endogenous switching regression)
I am doing a causal treatment effect analysis where my treatment variable is multivalued (it has 3 categories) and I am using the IPWRA framework for estimating ATT (teffects ipwra). To ensure robustness of my estimates while using this framework, I have to test that the conditional independence assumption (CIA) and overlapping assumption do hold. How can this be done based on the approach I am using for causal analysis?
2. With the same framework how can I justify that non-observable variables do not have much effect on my estimates? my reason for using IPWRA in a multivalued framework is because I did not have a valid instrument to control for selection on unobservables (Hence it is why I did not use multinomial endogenous switching regression)
