Open klintmane1 opened 3 years ago
Unfortunately, we've only implemented a regression forest in our grf module and not a classification forest. So currently there is no way to use a grf classifier for the propensity score.
However, many other forest classifiers would most prob perform equally well if not better in practice than a grf classifier. For instance, using xgboost or lightgbm or an sklearn RandomForestClassifier, or GradientBoostingClassifier, with hyperparameter tuning using cross valdiation would be equally good, potentially better.
See for instance our notebook here on how to do this: https://github.com/microsoft/EconML/blob/master/notebooks/ForestLearners%20Basic%20Example.ipynb
Hi!
This is more of a question rather than an issue. How would I use the generalized random forests to obtain propensity scores for my sample? Any examples or references?