Multilevel (hierarchical) modeling: what it can and can't do

Andrew Gelman

Introduction

Multilevel Modeling can be used for a variety of purposes, including prediction, data reduction, and causal inference from experiments and observation studies. Comparing to traditional regression, multilevel analysis are almost always an improvement. but to different degrees: for prediction, multilevel modeling can be essential, for data reduction, it can be uselful, and for causal inference it can be helpful. The Multilevle model is highly effective for precdictions at both levels but could be easily misinterpreted for causal inference.