Forecasting and Assessing Structural Change: the Case of Promotional Campaigns

Introduction We discuss and illustrate an important aspect of the modelling and analysis of time series data, namely structural change. We deal with this topic in the context of a general forecasting system employing dynamic (1985) extend the DLM class of models to include approaches for nonnormal and nonlinear models, and introduce the important practical concept of information discounting. West and Harrison (1989b) collects together the existing body of theory and lls in much additional mathematical background for the structure and analysis of the DLM, including an introductory discussion of work on numerical approaches to nonnormal and nonlinear dynamic models. Pole, West, and Harrison (1994) provides a more expository and applied introduction with computer software.