Bayesian Model Selection for High-Dimensional Ising Models, With Applications to Educational Data

TL;DR

This paper introduces a Bayesian method for high-dimensional Ising models that quantifies uncertainty more robustly than traditional penalized regressions, with applications to educational assessment data involving thousands of parameters.

Contribution

The paper develops a Bayesian approach for high-dimensional Ising models that overcomes limitations of existing methods by providing uncertainty quantification without strong assumptions.

Findings

Bayesian approach outperforms penalized regressions in robustness.

Method effectively handles models with thousands of parameters.

Simulation studies and real data applications validate the approach.

Abstract

Doubly-intractable posterior distributions arise in many applications of statistics concerned with discrete and dependent data, including physics, spatial statistics, machine learning, the social sciences, and other fields. A specific example is psychometrics, which has adapted high-dimensional Ising models from machine learning, with a view to studying the interactions among binary item responses in educational assessments. To estimate high-dimensional Ising models from educational assessment data, $\ell_1$-penalized nodewise logistic regressions have been used. Theoretical results in high-dimensional statistics show that $\ell_1$-penalized nodewise logistic regressions can recover the true interaction structure with high probability, provided that certain assumptions are satisfied. Those assumptions are hard to verify in practice and may be violated, and quantifying the uncertainty about the estimated interaction structure and parameter estimators is challenging. We propose a Bayesian approach that helps quantify the uncertainty about the interaction structure and parameters without requiring strong assumptions, and can be applied to Ising models with thousands of parameters. We demonstrate the advantages of the proposed Bayesian approach compared with $\ell_1$-penalized nodewise logistic regressions by simulation studies and applications to small and large educational data sets with up to 2,485 parameters. Among other things, the simulation studies suggest that the Bayesian approach is more robust against model misspecification due to omitted covariates than $\ell_1$-penalized nodewise logistic regressions.