Chunk #127 — Features and Pitfalls — Bias in Variable Selection and Variable Importance

Source

An introduction to recursive partitioning: rationale, application, and characteristics of classification and regression trees, bagging, and random forests.

Embedded

yes

Text

This bias is carried forward to ensembles of trees: Especially the variable importance can be biased when a data set contains predictor variables of different types (Strobl et al. 2007). The bias is particularly pronounced for the Gini importance, that is based on the biased Gini gain split selection criterion (Strobl et al. 2007), but can also affect the permutation importance. Only when subsamples drawn without replacement, instead of bootstrap samples, in combination with unbiased split selection criteria, are used in constructing the forest, the resulting permutation importance can be interpreted reliably (Strobl et al. 2007).