Customer Reviews

Time Series Analysis and Its Applications (Springer Texts in Statistics)

5 star:		(13)
4 star:		(6)
3 star:		(3)
2 star:		(3)
1 star:		(7)

 

 

Although a lot of books have been written on time series analysis, most of them just aren't very good. "Time Series Analysis and its Applications" is one of the better time series text books. It's not a brilliant book, but all of the other time series books that I have seen are worse.

This book covers all of the main areas of time series analysis such as ARIMA, GARCH and ARMAX models and spectral analysis and it does a pretty good job of it. Most of the explanations are clear enough for a beginner (with some statistical background) and are accompanied by worked examples (something which seems to be omitted in a lot of time series texts). Exercises are also provided at the end of each chapter, although no solutions are provided in the book (a colleague of mine informed me that the solutions are provided on the author's website, but that a large portion of these are either wrong or poorly explained).

Prospective purchasers of this book should be aware, however, that there are a number of errors throughout this book (corrections can be found on the author's website) and that although the title suggests that there are "R examples" in this book, these examples are few and far between and are not well explained. If you are looking for a manual for the R time series functions, then this is not the book for you.

I am a university lecturer and set this book as a supplementary text for an undergraduate statistics unit I teach, which includes a time series component. I believe that this is the best book available for this purpose. However, if you are a lecturer who is thinking of setting this book as a text for your class, please be aware of its limitations, and make sure that your students are also aware of them.

This is a modern book on time series analysis with many interesting and useful examples. It has a practical orientation much like Shumway's earlier book. The material has been tested in courses given by the authors at UC Berkeley and UC Davis. Good for both undergraduate and graduate level students. It covers most of the basics from both the time and frequency domain approaches. Although one reviewer suggests that it is light on theory compared to the Brockwell and Davis book, there is an adequate amount of theory presented which makes the level intermediate. It does require some advanced mathematics. Interesting topics not commonly found in competitor books include long memory ARMA models, the multivariate ARMAX models and their state space representation, applications of ARMAX models to longitudinal data analysis, bootstrapping state space models and the use of frequency domain time series methods applied to discriminant analysis, clustering and various other common multivariate statistical techniques. It also has a nice list of references. It definitely deserves 5 stars and possibly an oscar!

I am writing this review from a perspective of a graduate student taking a statistical time series class for the first time. Below you'll see a lot of good review of the book. But the question you should ask is how many of these reviewers used this book for their first time series classes?

Despite the claims of the authors, this is not a book for the beginners. It requires quite a bit of mathematical maturity and an in-depth knowledge of statistical methods. Here's a summary:

Disadvantages of the book:
1. It is a difficult and frustrating read.
2. Development of difference equations (fundamental tool in analyses of time series) is scattered everywhere and weak at best.
3. The material is not presented in a cohesive manner.
4. The author constantly relegates important theorems to the end of the chapter sections (which supposedly could be skipped on the first reading) and refers to these theorems in subsequent sections.
5. This book contains lots of typos.
6. Important results that must be discussed within the text material are left as exercises.
7. The notation is strange. Example: A random variable is universally represented by a capital letter such as an X. Author uses lower case letters to represent random variables.
8. The coverage of frequency domain is appalling. The author does a ghastly job of introducing Fourier Series and Transform. An entire chapter (chapter 3) is on frequency domain analysis. The question after reading the chapter is: so what???
9. No solutions or hints are provided so this book is practically useless for self-study.

Advantages:
1. It covers some recent developments in time series.
2. Its associated website has some decent data and S code.
3. It has a nice book cover.

There are plenty of other books better, or I should say much superior to this useless book:
1. Time Series Analysis by Hamilton
2. Introduction to Time Series and Forecasting by Brockwell
3. Applied Econometric Time Series by Enders (A bit outdated but very readable)
4. Analysis of Time Series by Chatfield (Lower level but a good book)

Conclusion: There are lots of other alternatives. This is a horrible book. It may be popular but I believe its popularity is due to good marketing and possibly good connections the authors may have.

If you expect this book to partly serve as a R manual for time series models, you'd be disappointed. In the last two hundred pages, namely chapter 6 "state space models" and chapter 7 "statistical models in the frequency domain", there are only mathematical formulas and no single line of R code. In chapter 5, the R code is only a few lines of calling garch function. Many of the R code in previous chapters are limited to "scan" and "plot" and one line of calling simple functions like arima. If you already learned time series from other books, you are better off reading CRAN website to find the relevant R function rather than buying this book

The book is OK but it falls behind other available texts at comparable or lower prices. I agree with others that the book is not the best introduction and neither a must-have rigorous reference. The main contribution is that it does account for some topics not typically found in most time series textbooks as mentioned in Dr. Chernick's review. The new edition of the classic by Box et al and the introductory text by Brockwell and Davis (ITSF) are muchl superior to Shumway and Stoffer in terms of introducing the core subject (ARIMA modelling) though not using R. If one wants R material (which by the way has powerful time series resources) than the book by Cryer and Chan does a much better job. If one wants more theory and technical detail, and also a solid introduction to multivariate methods, then the theoretical book by Brockwell and Davis (TSTM) and Hamilton's text are way better than this book. Applied economists wanting intro material should check Ender's applied text and engineers serious about time series cannot do better than owning Box et al and the (frequency domain) book by Percival and Walden. Statisticians and advanced readers can go to the two theoretical books I mentioned before.

As the title implies, the book is a text on time series analysis and its applications. It is a modern treatment of time series analysis with a slant toward applications. The applications are interesting and involve current topics such as global warming. The examples are broad in range, including data from various fields such as biology, economics, engineering, environmental science, and medicine. The book is interesting and accessible, and it provides an excellent introduction various aspects of the analysis of time series. The text covers both the spectral and time domains, including a thorough presentation of state-space models. The basic requirement for being able to understand most of the text is knowing the material that would be covered in introductory courses on regression and mathematical statistics.

The book has many interesting and "real" (as opposed to "toy") examples and, as the subtitle explains, many of the examples have associated R code. This makes for a positive experience because you can replicate the analyses. Accordingly, there is no guessing as to what was done to obtain the results of an example. It is completely wrong to say that "R is not relevant". But you do not have to take my word for it! Just go to the website for the text at StatLib and all the R code in the text is posted there. In addition, as the authors' state in the Preface (which is also available for viewing at the website for the text), R code for the state-space chapter (Chapter 6) is on the website for the text. There you will find code for the Kalman filter and smoothing algorithms, as well as the EM algorithm and some examples for maximum likelihood estimation. The website for the text also has a small tutorial for a quick start on using R to do time series analysis. The tutorial is great for a beginner.

At the end of the day, the text is not an R manual. It never says it is and I do not understand why anyone would think it should be an R manual. It is also not a manual for making wine and it will not help you train your dog. It is, however, an accessible modern introduction to time series analysis with many interesting examples that have associated R code. And, while you are learning time series analysis, you will also learn how to use R for analyzing time series.

I don't believe that the book deserves an oscar, but it is very good. I learned time series from Brockwell & Davis, and this book is less rigorous, but easier to understand and well motivated. The touches on the more modern stuff (long memory, bootstrap, etc) are slight, and a reader seeking info on these topics will need to turn to other sources, but an in depth treatment of long memory in an introductory book can not be expected.

I have not found any typos yet, and it reads very well.

Strongly recommended for anybody who wants to learn mathematical time series analysis ( I say mathematical, because it still has a decent amount of rigour, theorems, proofs, definitions, etc.), but does not intend to become a PhD in the field (in which case Brockwell & Davis might be a better choice).

As mentioned by some other reviewers, this book may be a good book in content, but it is very badly organized. The author references figures or equations from everywhere in the book. You have to go through chapters back and forth. Some important definitions are not clearly defined. They were just written into normal passages.

Dr Shumway and Dr Stoffer have produced a book upon time series analysis that will become an industry and academic standard. All those mathematical and diagnostic frighteners that have been sidestepped by many other authors have been introduced by the authors and used in such a simplifying way that students of all sciences, not only economics, will richly enjoy reading and putting into use. Garch,Bootstrapping, State-Space, Long-memory, if it is modern then it is covered in detail with plenty of top-notch examples. I give the book 5 stars and an Oscar.

This book is something... I've read it twice and still return to some chapter from time to time. It really requires patience and a strong mathematical background to get through some chapter but it gives you a knowledge and confidence in the modern time series statistics. The text is quite dense and concentrated but I like it.