search

jaredsmurray / bcf

Bayesian Causal Forests
39 stars 20 forks source link

Implementation of binary outcome #2

Closed arnaudfrn closed 4 years ago

arnaudfrn commented 4 years ago

Thanks so much for the work ! I read here (https://rdrr.io/cran/bcf/) that the bcf currently only works with continous outcomes. Any plans for an implementation for binary outcomes ?

Happy to work on a pull request if it can be easily implemented in R

jaredsmurray commented 4 years ago

Strictly speaking the implementation isn't difficult, but the interpretation is (tau is no longer the treatment effect on the outcome scale, but on a latent scale). While it's still a reasonable model it's difficult to calibrate by default, so it doesn't appear in this package right now.