'Poor overlap issue when estimating the Average (Conditional) Local Average Treatment by IV forest (instrumental_forest in the R package'grf')
(R package 'grf', generalized random forest)
My question is the output of average_treatment_effect() is NaN. I am using IV forest instead of conventional causal forest, and the estimand of my case is the Average (Conditional) Local Average Treatment (ACLATE):
average_treatment_effect(ivforest, target.sample = "all")
The warning message also indicated poor overlap, but I can't change to estimate ATE on the treated due to IV forest. The error message is as below:
Error in average_treatment_effect(ivforest, target.sample = "treated") : For any forest type other than causal_forest, the only implemented option is method=AIPW and target.sample=all
I am eager to know how to handle this condition. Because IV forest is very essential in my research. Thank you very much!
Warm regards, David
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|