search

grf-labs / grf

Generalized Random Forests
https://grf-labs.github.io/grf/
GNU General Public License v3.0
938 stars 250 forks source link

Questions Regarding Causal Forest Variable Types and "test_calibration" Interpretation #1409

Open Matt9617 opened 2 months ago

Matt9617 commented 2 months ago

Hi, Thanks for your work.

I am a student currently learning about causal forest and have some questions regarding the variable types and test_calibration Interpretation in causal forest.

  1. Does causal forest support scenarios where the treatment variable is continuous and the outcome variable is binary?
  2. The documentation states: "If Y.hat or W.hat = NULL, these are estimated using a separate regression forest." Does this mean I can skip predicting Y.hat and W.hat before setting up the causal forest and just use the default parameters?
  3. Is a significant value of differential.forest.prediction between 2 and 3 acceptable, and what is the implication of such a relatively high value?

Best Regards.

erikcs commented 2 months ago

Hi @Matt9617,

  1. Yes.
  2. That's correct.
  3. Stefan has a nice video lecture here on interpreting that kind of calibration exercise. You might find this recent grf feature, RATE, easier to interpret.