hdbayes: An R Package for Bayesian Analysis of Generalized Linear Models Using Historical Data

2025-06-24

AI Summary

In Bayesian regression, the absence of unified software support for prior modeling using historical data hinders methodological comparability across approaches. Method: We developed the R package *hdbayes*, implemented in Stan, to support multiple historical-data–driven priors—including power prior (POWER), meta-analytic predictive (MAP), and hierarchical Bayesian (HB) methods—within a standardized modeling interface and a unified MCMC inference framework. It enables flexible prior specification and cross-method comparison for generalized linear models. Contribution/Results: *hdbayes* substantially lowers the technical barrier to implementing Bayesian historical priors, enhances analytical reproducibility and methodological comparability, and improves empirical efficiency in evidence synthesis and regulatory science. By facilitating robust, transparent, and standardized Bayesian practice, the package advances principled integration of external evidence in statistical modeling and decision-making contexts.

Technology Category

• Machine Learning: Bayesian Learning
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

There has been increased interest in the use of historical data to formulate informative priors in regression models. While many such priors for incorporating historical data have been proposed, adoption is limited due to access to software. Where software does exist, the implementations between different methods could be vastly different, making comparisons between methods difficult. In this paper, we introduce the R package hdbayes, an implementation of the power prior, normalized power prior, Bayesian hierarchical model, robust meta-analytic prior, commensurate prior, and latent exchangeability prior for generalized linear models. The bulk of the package is written in the Stan programming language, with user-friendly R wrapper functions to call samplers.

Problem

• Lack of software for Bayesian analysis with historical data
• Inconsistent implementations hinder method comparisons
• Need user-friendly tools for generalized linear models

Innovation

• R package for Bayesian GLMs with historical data
• Implements multiple priors in Stan language
• User-friendly R wrappers for samplers