Simpson's paradox is one of many reasons why it’s important to evaluate your models on different slices of data.
Model 1 outperforms model 2 on group A and group B separately, but model 2 can still outperform model 1 overall.
Statistics is amazing.
Model 1 outperforms model 2 on group A and group B separately, but model 2 can still outperform model 1 overall.
Statistics is amazing.
