7.2 Summary of GRF

It seeks a generalized way to conduct causal inference under a non-parametric framework. GRF relies on random forest.

Methods developed to aid causal inference such as: $i)$ randomized controlled trial, $ii)$ comparison between treatment and control units under unconfoundedness assumption, $iii)$ difference-in-differences, and $iv)$ panel data methods; can fit into GRF framework. To do so, one needs to feed in the method-specific encoding into the GRF framework (to guide the splitting process).