Extendable priors for multivariate normal models that keep analysis simple and flexible.
This work presents an approach to broaden the Normal-Wishart family of priors so you can assign diagonal and other prior variances more freely to the parameters of a multivariate normal data-generating process. The authors show how to preserve analytical convenience while gaining much greater control over prior beliefs.
- Learn how to form extended conjugate distributions that stay the same functional form when updated with data.
- See how marginal distributions for the mean vector and the covariance-like parameter can be derived easily.
- Explore how these extensions support straightforward preposterior analysis and decision making.
- Understand the role of Bellman-type extensions and how they influence practical Bayesian modeling.
Ideal for readers of Bayesian statistics and multivariate analysis who want more flexible priors without sacrificing tractability.