Causal Random Forest

[Edgar Kausel]

Hi:

Is anyone aware or has anyone tried to implement Causal Random Forests to examine heterogeneous treatment effects in Stata?

As in here (code in R):
https://www.markhw.com/blog/causalforestintro

Thanks.
Edgar

[Adrien Bouguen]

Hi Edgar,

Any update on this question? I was wondering if someone is working on a stata adaptation of the R command GRF

https://grf-labs.github.io/grf/

Thanks

[Fabian Mierisch]

Hi Adrien and Edgar,

Please let me know if you found something. I would be interested, too.

Best,
Fabian

[Adrien Bouguen]

Hi Fabian,

I run the GRF on R and despite my basic skills I was able to run the analysis after a few days of work. This is worth it. I would recommand the tutorial of the Golub center from Stanford U. I imagine students from Susan Athey worked on this. It is really well made:

https://bookdown.org/halflearned/ml-ci-tutorial/

good luck !

[Adrien Bouguen]

I doubt anyone is working on this: this is an immense amount of work that was probably financed by a research fund at Stanford. I don't see anyone (except Stata corp) starting such endeavor.

Also, I am under the impression that Stata is not the right plateform to run very large random forest algorithm. I tried the random_forest command on 540 variables and the computer took for ever to find a solution while the equivalent command in R was much faster. I don't know why but I imagine that stata has not been optimized for ML. I guess the recent interagration of Python will help but I guess the solution will be to have a R integration into Stata. I know that some people have been talking about this.

Anyone has more info about the R integration? The python integration? Or the difference in terms of architecture between R and Stata?