-
I tried
`X=matrix(rnorm(100), 100,1)
Y=t(t(rnorm(100)))
W=t(t(rbinom(100,size=1, prob=0.5)))
forest = causal_forest(X,Y,W)
average_partial_effect(forest)`
And the following alert comes out
…
-
**Description of the bug**
The function `compute_debiased_error` actually returns raw estimates of the error, without debiasing.
The issue is in [this line](https://github.com/grf-labs/grf/blob/9c…
-
I installed successfully gradient-forest package on my Mac, but was getting an error while running test example provided in the repository documentation.
> tau.forest = causal.forest(X, Y, W)
Err…
-
For the causal forest algorithm to work, there must be sufficient overlap among similar samples in terms of their treatment. Currently when given data with very little overlap, `causal_forest` may fai…
-
I think this package is really interesting. It would be nice if there was a function to do a varImpPlot, similar to what is available in the randomForest package. Right now there is nothing I am aware…
-
I'm generating simple linear data (in logit space), but the predicted TEs from `causal_forest` seem to be uncorrelated with the true oracle TEs. I would expect much better results than I'm currently g…
kcrum updated
5 years ago
-
Hello,
I am just wondering why the honesty parameter is set to TRUE by default ?! And do you think for next releases we could use grf with caret ?
Thank you
-
**Description of the bug**
There are certain scenarios in which having a nonzero `imbalance.penalty` causes most trees to split very few times. This obviously causes the performance of the prediction…
-
Suppose I estimate a `regression_forest` with matrices `Y` and `X`, and I construct a matrix of the same shape as `X`, call it `Z`, but with possibly perturbed values. Is it possible to recover `E[y|…
-
While migrating from the previous -causalTree- package to -grf-, I noticed a curious case for -grf- when running the following causal_forest example. When I expand the range of true treatment effect (…