Bayesian change-point analysis in hydrometeorological time series. Part 2.Comparison of change-point models and forecasting

This paper provides a methodology to test existence, type, and strength of changes in the distribution of a sequence of hydrometeorological random var iables. Unlike most published work on change-point analysis, which consider a single structure of change occurring with certainty, it allows for the c onsideration in the inference process of the no change hypothesis and vario us possible situations that may occur. The approach is based on Bayesian mo del selection and is illustrated using univariate normal models. Four univa riate normal models are considered: the no change hypothesis, a single chan ge in the mean level only, a single change in the variance only, and a simu ltaneous change in both the mean and the variance. First, inference analysi s of posterior distributions via Gibbs sampling for a given change-point mo del is recalled. This scientific reporting framework is then generalized to the problem of selecting among different configurations of a single change and the no change hypothesis. The important operational issue of forecasti ng a future observation, often neglected in the literature on change-point analysis, is also treated in the previous model selection perspective. To i llustrate the approach, a case study involving annual energy inflows for ei ght large hydropower systems situated in Quebec is detailed. (C) 2000 Elsev ier Science B.V, All rights reserved.