APS

29th APS Annual Convention · 2017

What to Do with Outliers in Bayesian Analysis

Boston, MA · May 2017

Poster Session · Methodology

  • Bommae Kim
    Federal Reserve Bank of Kansas City
  • Xin Tong
    University of Virginia

Abstract

A Monte-Carlo simulation study examined different methods to treat outliers in Bayesian analysis. Robust estimation using a t-distribution and rank-based inverse normal transformation (RIN) showed superior performance over no-treatment, outlier deletion by Cook’s D, and Box-Cox transformation.

Quantitative

← Poster Session <span>III</span>