Supplementary Files - Janitza, Tutz and Boulesteix (2015)
Random Forest for Ordinal Responses: Prediction and Variable Selection
Contact: Silke Janitza, email: email@example.com
Implementation of variable importance measures for ordinal response
The novel variable importance measures (VIMs) introduced in the paper are implemented for the statistical software R for the package party. Make sure that the party package is installed before sourcing the functions. The implementations for the VIMs are mainly based on the code for function varimp contained in the party package.
- Click here for the code implementing the RPS-based VIM (function 'varimpRPS'), the MAE-based VIM ('varimpMAE') and the MSE-based VIM for classification and ordinal regression trees ('varimpMSE').
- Please also see our tutorial for an illustration on the application of the functions, and the R-code used in the tutorial to demonstrate the application of the VIMs.
Reproducible files download
- The ZIP-file (596 KB) contains all R scripts to perform the simulation studies in the Technical Report 174, Department of Statistics.