Nonlinear Time Series Analysis

Rating:
Summary: Good for both beginners and advanced practitioners
Review: In my search for good material on time series analysis, I have come across many books packed with information, yet so dry as to make them unreadable (readers of Hamilton's "Time Series Analysis" will know what I mean - Amazing book, but unreadably boring).

Kantz and Schreiber do not suffer from that all too common problem. They write clearly and in a very readable style. Their use of real-world datasets and numerous (though not overwhelming) charts makes their work quickly accessible even to beginniners in the field. They provide enough mathematical formalisms to make use of what they present, but not so many as to require a PhD in math to follow the flow of the text. For more advanced readers, they cover a wide range of topics useful both for analysis and for forecasting. Chapter 12, in particular, opened me to a whole world of new techniques.

As my one negative comment on this book, I would have liked that same chapter 12 fleshed out more, to the point that I would buy a follow-up book covering nothing but an elaboration on that single chapter.

If you have an interest in time series analysis and forecasting, and have grown tired of dry material that provides nothing more than yet another extension to ARIMA or Kalman filtering, you will love this book.

Rating:
Summary: Excellent for practitioners
Review: This book provides an excellent overview of chaos theory concepts applied to time series analysis. First part constitutes a good tutorial on chaos theory and its implications on time series analysis while the second part discusses in detail aspects of time-series related chaos theory concepts (with an historical perspective of the related research). Time series analysts will certainly benefit from it thanks to its balanced exposition of issues of chaos theory concepts for non-infinite data sets...

However, the only drawback is that it essentially deals with deterministic systems, not stochastic ones. But if you gathered your data on a physical system, it's OK.