jaredsmurray / bcf

Bayesian Causal Forests
40 stars 20 forks source link

A few enhancement requests #13

Closed schwa021 closed 2 years ago

schwa021 commented 3 years ago

Thanks for the great software (and help via personal email. I should have posted here instead. mea culpa)

Bugs

Minor bug - CRAN version does not return posterior distributions for mu. Note that the dev version does do this! Minor bug - The dev version works like a charm, but seems to run extremely slowly on my device (5x or worse slower than the CRAN version)

Enhancement Requests [UPDATE - I JUST NOTICED THE DEV VERSION DOES HAVE A PREDICT METHOD WHICH WORKS GREAT]

It would be really helpful to have a "predict" function (or, perhaps even better, just return the actual BART). This would allow straightforward testing on independent samples for out-of-sample performance assessment, as well as implementation for use in "real-world" settings (e.g. providing pre-treatment predictions to M.D.s in an actual clinical setting - which is something I would like to do).