David Afshartous

Overview

Statistician and educator; primary author of Afshartous and de Leeuw (2005), a systematic comparison of three prediction rules for hierarchical linear models published in the Journal of Educational and Behavioral Statistics.

Key Contributions / Features

• Afshartous and de Leeuw (2005) — "Prediction in Multilevel Models," Journal of Educational and Behavioral Statistics 30(2): 109–139. Proved that the multilevel (shrinkage) predictor equals the Bayes rule under squared error loss (Proposition 1.1); established that $n_j \to \infty$ collapses it to OLS and $\rho \to 0$ collapses it to the prior predictor; derived bias–shrinkage formula for underestimated Level 2 variance; simulation over 30,000 datasets showing multilevel dominates in 24/25 $J \times n_j$ design cells.

Related

• Mixed-Effects Regression
• Bayesian Hierarchical Model
• Random Coefficient Model
• Jan de Leeuw
• Afshartous-de Leeuw (2005)