MacDonald L., Zucchini W. Hidden Markov and Other Models for Discrete-valued Time Series

Chapman & Hall / CRC Press, 1997. — 256 p.Discrete-valued time series are common in practice, yet methods for their analysis are not well known. Many university courses on time series pay little or no attention to methods and models specifically designed for discrete-valued series, although the list of discrete-valued models proposed in the research literature is by now a long one. These models have not found their way into the major statistical packages nor, we believe it is fair to say, into the repertoire of most applied statisticians. The reason may be that there is no well-known family of models that are structurally simple, sufficiently versatile to cater for a useful variety of data types, and readily accessible to the practitioner.
We have two main objectives in this monograph. Firstly, we wish to provide a summary of the models that have been proposed and of such data-analytic methodology as has been developed for these. Secondly, we wish to describe in detail the class of hidden Markov models. In our opinion this class of models possesses enough of the desirable properties n1entioned above to make it worthwhile for the applied statistician to become familiar with them. The two parts of the book reflect these two separate objectives, and can be read independently.
Our intended readership is primarily applied statisticians who meet d1screte-valued time series in their work, and who need to know what appropriate methodology is available and how to apply it. The book is also intended for statistical researchers who wish to contribute to the further development of the subject. Finally, some of the material might profitably be included in a graduate or even undergraduate- course on time series analysis.Survey of models
A survey of models for discrete-valued time series
Hidden Markov models
The basic models
Extensions and modifications
Applications