GATEs - true individual effects (ite) questions #303
-
|
Hi all, I apologize if this is a basic question. I am using the DML package to find the Group Average Treatment Effect (GATEs) for the PLR model using my local data. However, I am unsure how to extract the ITE (true individual effects) using the "effects" key. Can anyone help me identify the attributes or explain how to extract them? Secondly, when should we use the Best Linear Predictor model (bl_model)? Is it used only in the IRM model, or can it also be used for the PLR model? Should we fit it before or after GATE CI? Thank you in advance for your assistance.
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi @jinny-nguyen, thanks for the questions. The true ITEs for the GATE-Example are only available due to the data being generated by make_heterogeneous_data. These are not part of the estimated object and only used to compare underlying effects to the estimated ones in the example. The BLP is also used in the PLR-model, but this is handled via the |
Beta Was this translation helpful? Give feedback.
Hi @jinny-nguyen,
thanks for the questions.
The true ITEs for the GATE-Example are only available due to the data being generated by make_heterogeneous_data. These are not part of the estimated object and only used to compare underlying effects to the estimated ones in the example.
The BLP is also used in the PLR-model, but this is handled via the
gate(orcate) method. The methods internally fit a best linear predictor in a specific form. You can find a documentation here CATE-PLR-Model. Thegatemethod does something similar to a basis functions of indicators.