Iain Pardoe

Overview

Iain Pardoe is a statistician at the University of Oregon (Lundquist College of Business) at the time of his collaboration with Gelman. He specializes in applied regression analysis and hierarchical modeling.

Key Contributions

• Gelman-Pardoe (2004): Co-developed with Andrew Gelman the level-specific R² (eq. 7, proportion of level-m variance explained by predictor μ) and pooling factor λ (eq. 8, fraction of residual variance reduced by hierarchical shrinkage). Established the Bayesian adjusted R² with correction factor $(n-3)/(n-p-2)$ and connected λ to the classical Bühlmann credibility weight ω for balanced one-way hierarchical models. Applied measures to the Minnesota radon study (919 houses, 85 counties), finding $R^2_\beta = 0.83$ and $\lambda_\beta = 0.97$ at the slope level. See Bayesian Hierarchical Model.

Related

• Andrew Gelman
• Bayesian Hierarchical Model
• Gelman-Pardoe (2004) Bayesian Measures of Explained Variance and Pooling in Multilevel (Hierarchical) Models