Model choice versus model criticism

TL;DR

This paper discusses new perspectives on Approximate Bayesian Computation (ABC) and Bayesian model criticism, challenging traditional Bayesian model choice methods by examining prior influence, model assessment, and the interpretation of error.

Contribution

It critically analyzes recent developments in ABC and Bayesian model criticism, highlighting issues and conceptual challenges in these approaches.

Findings

Highlights prior influence on model choice

Examines issues in model assessment and criticism

Questions the meaning of error in ABC methods

Abstract

The new perspectives on ABC and Bayesian model criticisms presented in Ratmann et al.(2009) are challenging standard approaches to Bayesian model choice. We discuss here some issues arising from the authors' approach, including prior influence, model assessment and criticism, and the meaning of error in ABC.