Efficient Sampling for Gaussian Linear Regression With Arbitrary Priors

P. Richard Hahn of Arizona State University developed a slice sampler for Bayesian linear regression models. The sampler is designed to be faster than custom implementations for models with many regressors and can work with any prior with an evaluable density function. It is intended for investigating new shrinkage priors without requiring custom sampling algorithms.

Use Cases

• Benchmarking sampling efficiency for high-dimensional regression based on the described sampler's speed advantage.
• Investigating properties of new shrinkage priors based on the sampler's compatibility with arbitrary prior densities.
• Comparing effective sample size per second against alternative sampling approaches for linear regression.

Strengths

• The sampler is described as faster than many custom implementations when the number of regressors is large.
• The sampler can be used with any prior whose density function can be evaluated up to a normalizing constant.
• The method reportedly produces better effective sample size per second than common alternative approaches.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

P. Richard Hahn, Arizona State University

Collection Method

Methodological development for statistical sampling.