Online Program Home
My Program

Abstract Details

Activity Number: 260 - SPEED: Topics in Bayesian Analysis
Type: Contributed
Date/Time: Monday, July 30, 2018 : 3:05 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #332975
Title: Pseudo-Marginal Markov Chain Monte Carlo via Random Riemann Sums for Stochastically Scaled Gaussian Vectors
Author(s): Patrick Muchmore*
Keywords: Markov chain Monte Carlo; pseudo-marginal; density estimation

A wide variety of probabilistic models can be encapsulated in the framework of a Gaussian vector stochastically scaled by a positive random variable. Although conceptually straightforward, in all but a few cases the resultant random vectors posses density functions which are either computationally intractable or completely unknown. In this work we consider the problem of density estimation for these random vectors. We first show how the problem can be reduced to sampling from the positive real numbers and evaluating the multivariate Gaussian density, yielding a standard Monte Carlo estimator. We also demonstrate how significantly higher accuracy can be achieved, at roughly the same computational cost, by employing the theory of random Riemann sums. Both of these approaches yield unbiased estimates of the likelihood; therefore, they enable exact inference despite an intractable, or even unknown, likelihood function by appealing to the recently developed theory of pseudo-marginal Markov chain Monte Carlo.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program