Small Sample Methods for Multilevel Modeling: A Colloquial Elucidation of REML and the Kenward-Roger Correction

Studies on small sample properties of multilevel models have become increasingly prominent in the methodological literature in response to the frequency with which small sample data appear in empirical studies. Simulation results generally recommend that empirical researchers employ restricted maximum likelihood estimation (REML) with a Kenward-Roger correction with small samples in frequentist contexts to minimize small sample bias in...