Introduction to Time-Series Modelling, by Genshiro Kitagawa, Boca Raton, Chapman & Hall/CRC, 2010, xxiii+289 pp., \pounds49.99 or US$79.95 (hardback), ISBN 978-1-58488-921-2
This book contains 16 compact chapters, each 10–15 pages long, written in a rather small font size, rapidly contributing to the reader's eye fatigue. A larger font size and more pages would have been appreciated. All these chapters end with a small “problems” section where the author suggests some exercises. Corresponding (brief) solutions are given at the end of the book, from pages 263 to 276. A bibliography section is provided at the end of the book with a relatively small number of references (89).
The first five chapters have to be considered as introductory ones: how the book is organised, the covariance function, the power spectrum and the periodogram, Kullback–Leibler and Akaike information criteria, and the least-squares method. The time-series modelling part of the book actually starts at Chapter 6 with the introduction of ARMA models. The corresponding estimation of model parameters is addressed in Chapter 7 for the AR model and in Chapter 10 for the ARMA model. The order of some chapters is sometimes intriguing. For instance, Chapters 8 and 13 are devoted to particular kinds of autoregressive (AR) models, i.e.locally stationary and time-varying coefficient AR models. Why not put them together, one after another? The same remark holds for Chapters 9 and 14, both devoted to the state-space model (and more precisely to filtering and smoothing algorithms). Chapters 11 and 12 are econometric/financial chapters where trends and seasonal adjustment have to be taken into account. Finally, I consider that the two last chapters are a bit off the main topic of the book: Chapter 15 addresses the sequential Monte Carlo filter while Chapter 16 is devoted to how to simulate random numbers.
After having read all these chapters, the reader will probably be disappointed to not have found any material related to some other important time-series models like the integrated moving average (IMA) or the exponentially weighted moving average (EWMA) models that have applications in numerous fields (statistical process control for instance), extended autoregressive moving average model like the autoregressive moving average with external input (ARMAX) model or even generalized autoregressive conditional heteroskedasticity (GARCH) models that are usually used to model the volatility processes of financial time series.
In conclusion, I am a bit disappointed by this book (and I tend to not recommend it) for the following reasons:
| • | It tends to cover a large spectrum of topics but without addressing them very deeply. | ||||
| • | The order of the chapters is sometimes odd and I would have preferred either a clear topic order (i.e.chapters with the same kind of topic close to each other) or, even better, a division of the book into three clear parts: (1) introductory chapters, (2) fundamentals of time series and (3) advanced topics. This could help the reader to know what is important and what is not. | ||||
| • | Some standard modern topics are totally missing, like IMA, EWMA, ARMAX, GARCH time series. | ||||