Robust Bayesian Inference via Coarsening: Simulated and Real Data Examples

Jeffrey W. Miller from Harvard University introduces a novel Bayesian inference method that improves robustness to model misspecification. The approach conditions on the model generating data close to the observed data, approximated by tempering the likelihood. The paper illustrates the method with real and simulated data using mixture models and autoregressive models of unknown order.

Use Cases

• Implementing robust Bayesian inference based on the described coarsened posterior and likelihood tempering.
• Analyzing mixture model data with improved robustness to small departures from the model assumption.
• Fitting autoregressive models of unknown order using the coarsening framework for distributional closeness.
• Studying theoretical properties of Bayesian procedures under model misspecification as derived in the paper.

Strengths

• Method developed by an author affiliated with Harvard University.
• Approach can be implemented via standard algorithms and offers analytical solutions with conjugate priors.
• Includes illustrations with both real and simulated data.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, column names, and file formats are unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the examples chosen for the paper's illustrations.

Provenance

Source

Jeffrey W. Miller, Harvard University

Collection Method

Likely contains data generated for methodological illustration in an academic paper.