Dirichlet Process Mixtures of Block g Priors for Linear Models

A research paper introducing Dirichlet process mixtures of block g priors for model selection and prediction in linear models. The work was authored by Anupreet Porwal and published on figshare in June 2026. It includes empirical performance investigations using various real and simulated datasets.

Use Cases

• Comparing model selection performance against traditional mixtures of g priors based on the described differential shrinkage method.
• Implementing the described Markov chain Monte Carlo algorithm for posterior inference in linear models.
• Investigating the avoidance of the conditional Lindley paradox in Bayesian model selection as discussed in the paper.

Strengths

• The paper includes empirical performance tests on various real and simulated datasets.
• The associated Markov chain Monte Carlo algorithm is described as requiring minimal ad-hoc tuning.
• The prior is shown to be consistent and to avoid a specific statistical paradox highlighted in prior literature.

Limitations

• The dataset is a 2.3 MB PDF document; the underlying data used for the empirical investigations is not directly provided.
• Column-level documentation for any accompanying data is absent; field semantics must be inferred from the paper.
• Row count for any empirical data is unknown, which may limit suitability assessment.

Provenance

Source

Anupreet Porwal via figshare.

Collection Method

Theoretical development and empirical evaluation, method described in the paper.

Freshness

Last updated 2026-06-02 15:30:43.