🤖 AI Summary
Standardized power priors are hindered in widespread Bayesian applications due to the intractability of their normalizing constants. This work proposes a general Markov chain Monte Carlo (MCMC) strategy that circumvents explicit computation of the marginal likelihood by exploiting a functional identity between the marginal distribution of the power parameter and the normalizing constant. The approach integrates seamlessly into mainstream Bayesian software platforms such as Stan, JAGS, and NIMBLE, substantially lowering the computational barrier to implementation. Empirical evaluations across multiple case studies demonstrate that the method is both practical and efficient, offering a viable pathway for routine use of standardized power priors in Bayesian inference.
📝 Abstract
The normalized power prior provides a principled framework for incorporating historical data into Bayesian inference while preserving coherence, but its routine application has been hindered by the need to evaluate an intractable normalizing constant function $C(a_{0})$. Existing approaches typically rely on model-specific marginal likelihood calculations, numerical integration over grids, or auxiliary sampling schemes implemented outside standard Bayesian software. In this paper, we present a computational perspective that exploits a simple and underutilized functional identity: under the unnormalized power prior, the marginal distribution of the power parameter is proportional to the normalizing constant of the normalized power prior. Leveraging this relationship, we propose a sampling-based strategy to approximate the normalized power prior using output from generic Markov chain Monte Carlo (MCMC) algorithms. The resulting approximation can be implemented entirely within all general Bayesian software packages (such as PROC MCMC, BUGS, JAGS, Stan, or NIMBLE), without requiring explicit marginal likelihood evaluation. Several illustrative examples demonstrate the practicality of the approach and highlight its potential to facilitate the routine use of normalized power priors in applied settings.