Unconditional large deviation principles for Dirichlet posterior and Bayesian bootstrap
Shui Feng · The Annals of Applied Probability · 2025
The unconditional or annealed large deviation principles are established for the Dirichlet posterior and the Bayesian bootstrap. The rate functions are identified explicitly, which provide new measurements of divergence between probabilities. As applications, we study the asymptotic efficiencies of the Dirichlet posterior mean and the Bayesian bootstrap mean.