A Predictive View of Inference About the Structural Change in Linear Models

In this paper, using some idea of statistical diagnostics such as Geisser （1985,1987）, we analyze structural changes in linear models from Bayesian predictive point of view.In this paper,we present two new methods basing on the Conditional Predictive Discordancy Diagnostics and Kullback-Leibler divergence respectively,to detect changes in regression parameters.