[R] sums of absolute deviations about the median as split function in rpart

[Sina Muster]

Dear R community,

as stated in Breiman et.al. (1984) and De'Ath & Fabricius (2000) using
sums of absolute deviations about the median as an impurity measure
gives robust trees.

I would like to use this method in rpart.

Has somebody already tried this method in rpart? Is there maybe already
a script available somewhere?

I am aware of the possibility to define usersplits myself with eval,
split and init. As an R beginner, though, I would like to check the
existing experience first.

Thanks
Sina


Writing a thesis on "The effects of land-use change on vegetation
patterns in a Southern-Alpine valley (Switzerland)"

University of Potsdam, Germany
Department of Geoecology