Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 27 - SPEED: Statistical Learning and Data Challenge Part 1
Type: Contributed
Date/Time: Sunday, August 7, 2022 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #322624
Title: Fast Bayesian Estimation for Ranking Models
Author(s): Michael Pearce*
Companies: University of Washington
Keywords: rank data; Mallows model; preference aggregation; peer review

The Mallows' model is a common location-scale family of ranking distributions. In the frequentist setting, estimation is challenging as the parameter space is high-dimensional and discrete, leading to slow, if not intractable, estimation. We propose a Bayesian procedure that allows for fast estimation of the Mallows' model via a posterior that can be expressed as a finite mixture of continuous parameter models. We demonstrate how prior distributions may be chosen to align Bayesian posteriors with those from the frequentist setting, allowing for substantial increases in estimation speed. In addition, we show how our procedure may be applied to various extensions of the Mallows' model in the literature. The model is applied to real grant panel review data to demonstrate fast preference aggregation and how to estimate its uncertainty.

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

Back to the full JSM 2022 program