Chunk #32 — Methods — Power calculation

Source

Random forest versus logistic regression: a large-scale benchmark experiment.

Embedded

yes

Text

For example, the required number of datasets to detect a difference in performances of δ=0.05 with α=0.05 and 1−β=0.8 is Mreq=32 if we assume a variance of σ2=0.01 and Mreq=8 for σ2=0.0025. It increases to Mreq=197 and Mreq=50, respectively, for differences of δ=0.02.