Nonparametric predictive inference for discrete data via Metropolis-adjusted Dirichlet seq

A research article presenting a novel Bayesian predictive approach for discrete distributions, focusing on count data. The methodology, developed by Davide Agnoletto, avoids computational disadvantages of traditional mixture models using a Metropolis-adjusted Dirichlet sequence model. The article is available as a 9.8 MB package on figshare under a CC-BY-4.0 license.

Use Cases

• Conducting Bayesian inference on unknown discrete distributions based on the described predictive approach
• Modeling count data using the Metropolis-adjusted Dirichlet sequence model to avoid traditional mixture model disadvantages
• Extending predictive methodology to multivariate, regression, and binary data cases as mentioned in the description
• Applying the asymptotic Gaussian approximation for inference on count distributions as illustrated in the article

Strengths

• Article is 9.8 MB in size, providing detailed methodological content
• Methodology is explicitly designed to address computational challenges with traditional mixture models
• License is CC-BY-4.0, permitting broad reuse and adaptation

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

figshare

Collection Method

Research article presenting a novel statistical methodology.

Time Range

Publication date is unspecified; last platform update was 2026-05-01.

Freshness

Last updated 2026-05-01 14:56:35; freshness should be verified

Geography

Geographic coverage is unspecified.