New R Package in Staburo: shadowVIMP – A Variable Selection Method for Random Forests 

At Staburo, we’re committed to developing solutions to statistical challenges that go beyond standard methods – supporting both client projects and scientific progress. 

A recent example of this commitment is shadowVIMP, a novel variable selection method based on random forest importance measures and permutation testing. Developed in-house by our colleagues Tim Müller, Hannes Buchner, and Maria Blanco, together with Dr. Roman Hornung, Ludwig Maximilian University of Munich, Germany, the method was met with great interest at conferences in Vienna and Berlin. 

ShadowVIMP adresses key challenges in high-dimensional data analysis, especially when variables are correlated – a known issue of the random forest variable importance measure. By preserving the correlation structure in the permutations and adjusting for multiple testing, it aims to offer more reliable and interpretable variable selections for exploratory analyses. 

We’re pleased to share that the method has been implemented as an open-source R package by Oktawia Miluch and Tim Müller, which is now available on CRAN. 

This work is one example of how we invest in methodological development to challenges we face in our project work. If you’re interested in collaborating on applied methodology – or in joining a team that values your statistical expertise – we’d love to hear from you. 

• Explore the package on CRAN 

• Interested in working with us? Contact us via the button below.