Abstract
A Bayesian approach is presented for influential observations and predictive discordancy diagnostics. Tests of significance based on predictive distributions are devised to test the discordancy of an observation. Examples illustrating the methods are provided.

This publication has 17 references indexed in Scilit: