|
Activity Number:
|
37
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305215 |
|
Title:
|
A Bayesian Approach to Detecting Outliers in Circular Data
|
|
Author(s):
|
Karen D.S. Young*+ and Lawrence I. Pettit and Nalaiyini Sothinathan
|
|
Companies:
|
University of Surrey and Queen Mary University of London and Queen Mary University of London
|
|
Address:
|
Department of Mathematics, FEPS, Guildford, GU2 7XH, United Kingdom
|
|
Keywords:
|
Bayesian methods ; Von Mises distribution ; Gibbs sampling
|
|
Abstract:
|
Data consisting of angles on a circle are often modeled by the Von Mises distribution with mean and concentration parameters. Damien and Walker (Can. J. of Statist., 1999) describe how to use Gibbs sampling to simulate samples from the posterior distributions of the parameters of such a distribution based on a proper prior distribution. We firstly extend their method to allow for fitting a possible outlier. Then we discuss how to evaluate a Bayes factor for deciding whether an observation is indeed an outlier. We consider the sensitivity of the inference to the choice of prior parameters. We also discuss particular problems which arise because of the nature of circular data and inferences can depend on the choice of origin. We illustrate the methods using the well known data set relating to the homing ability of the northern cricket frog and compare our results with previous analyses.
|
