Online Program Home
My Program

Abstract Details

Activity Number: 409
Type: Invited
Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #318403 View Presentation
Title: Fiducial Inference: Fisher's Big Blunder or Big Bang?
Author(s): Keli Liu and Xiao-Li Meng*
Companies: Stanford University and Harvard
Keywords: Distributional Estimator ; Estimating Equations ; Structural Equations ; Stochastic Algebra ; Parallel Computation ; Prior-free Posterior

Estimating equations are a popular avenue for creating point estimates. Fisher dared to also obtain a distributional inference by solving a system of "structural" equations. In terms of its original intent-a prior free posterior-Fiducial has been historically regarded as Fisher's Big Blunder. However, the era of Big Data encourages us to look at Fiducial from a new perspective, as an algorithmic revolution: stochastic algebra replaces Markov Chain Monte Carlo in computing posterior inferences. The trick to achieving this algorithmic jujitsu lies in converting the usual problem of inferring signals (parameters) into its dual form of predicting noise. Considering a problem in its dual form often suggests surprisingly simple algorithms for traditionally difficult tasks, e.g., parallelizable posterior computation and robust Bayesian analysis. We therefore have much reason to revisit Fiducial, perhaps turning Fisher's Big Blunder into a Big Bang.

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

Back to the full JSM 2016 program

Copyright © American Statistical Association