Using Simulation to Solve Problems

Abstract

TIME SERIES Analysis has received a lot of attention lately. When we are not being bombarded with leaflets announcing the absolutely last Bayes-Kalman forecasting event (rather like the end of the world is nigh announcements), we are treated to an acrid exchange on the merits of Box-Jenkins methods. For those of you who want a rest from all this high-powered argument, Montgomery and Johnson might make a welcome change with its affection Least Squares and its many worked examples. The main part of the book covers regression models, moving averages, exponential smoothing, direct smoothing, smoothing seasonal data, analysis of forecast errors, and adaptive control. None of the topics is treated exhaustively. A model is described, there are a few pages of algebra and then a lengthy worked example. The authors make no attempt to justify the choice of particular models the time series in their examples. They are too busy showing the reader how to put the numbers in the formulae to have time to discuss with him why they are using those formulae. They see their book partly as a reference someone who has already decided what model he wants to use but needs to check up on the mechanics of it. That seems to me to be passing the buck and certainly does not help in the last two chapters on Box-Jenkins models and Bayesian methods in forecasting. There just is not enough room to expound BoxJenkins methods in one chapter, and palming off parameter estimation on to a program that may be purchased for a modest charge (not from the authors I hasten to add), may relieve the lack of space but does not assist in developing the working knowledge of the technique which the dust jacket promised. The final chapter is a short introduction to Bayesian ideas, but in common with the rest of the book there is more mathematical manipulation than discussion of concepts. Throughout the book the mathematical approach is unsatisfactory. Formulae are presented, sometimes proved, sometimes not. The authors have taught the subject several years and are professional consultants. Surely there is much of their experience they could have passed on to make the book both more useful to the practitioner and more exciting the student.