Do we really need the special code in regr.randomForest for se estimation if we have that in BaggingWrapper?

[mlr-org]

Do we really need the special code in regr.randomForest for se estimation if we have that in BaggingWrapper?

[berndbischl]

We need to update the code and take that here into account

https://github.com/zmjones/edarf

[zmjones]

most of my code comes from randomForestCI. their variance estimator requires as input a matrix of bootstrap weights and the tree predictions as well, which neither rfsrc nor cforest provide. There is code to do both of those things in edarf. randomForest doesn't return a proper bootstrap weights matrix, only giving you an indicator of whether an observation was in the bootstrap sample for a tree, not the number of times it appears. Wager has also fixed this but I don't think it has been merged yet. I assume it will be at some point though.

I will look at integrating some of this stuff as soon as I get a chance!

[zmjones]

outdated/superseded by discussion in #595. will open an issue for the baggingwrapper se estimators when that is closed

[berndbischl]

ok

[berndbischl]

in general i would like to have the issues stay really open until they are solved, but what you say seems ok to me