General Bayesian Updating and the Loss-Likelihood Bootstrap

TL;DR

This paper introduces the loss-likelihood bootstrap, a method for Bayesian updating that extends the weighted likelihood bootstrap to parameters minimizing expected loss, providing a nonparametric interpretation and calibration technique.

Contribution

It derives the loss-likelihood bootstrap as an exact method under a Bayesian nonparametric framework, connecting it to general Bayesian updating and Fisher information calibration.

Findings

The loss-likelihood bootstrap accurately approximates Bayesian posteriors for loss-based parameters.

The method effectively calibrates general Bayesian posteriors without a full probabilistic model.

Demonstrations show practical applicability across various examples.

Abstract

In this paper we revisit the weighted likelihood bootstrap, a method that generates samples from an approximate Bayesian posterior of a parametric model. We show that the same method can be derived, without approximation, under a Bayesian nonparametric model with the parameter of interest defined as minimising an expected negative log-likelihood under an unknown sampling distribution. This interpretation enables us to extend the weighted likelihood bootstrap to posterior sampling for parameters minimizing an expected loss. We call this method the loss-likelihood bootstrap. We make a connection between this and general Bayesian updating, which is a way of updating prior belief distributions without needing to construct a global probability model, yet requires the calibration of two forms of loss function. The loss-likelihood bootstrap is used to calibrate the general Bayesian posterior by matching asymptotic Fisher information. We demonstrate the methodology on a number of examples.