# Books for self-studying time series analysis?

I started by Time Series Analysis by Hamilton, but I am lost hopelessly. This book is really too theoretical for me to learn by myself.

Does anybody have a recommendation for a textbook on time series analysis that's suitable for self-study?

• I think should be a community wiki question. – Rob Hyndman Jan 3 '12 at 8:41
• Could you provide a little bit more details on what are your particular needs: academic (scientific, PhD), practical (model building, engineering, programming), level of disaggregation (macro, micro, panel data), field of application (microeconomics, macroeconomics, finance, physical sciences), may be some other details you feel are relevant. – Dmitrij Celov Jan 4 '12 at 12:20
• I have always been a big fan of The Analysis of Time Series by Chris Chatfield – kaybenleroll Mar 3 '15 at 23:35
• I have a strong personal bias for amazon.co.uk/Time-Series-Analysis-Univariate-Multivariate/dp/… sorry @Taylor he does not treat the idea of Intervention Detection which is critical in identifying useful models. – IrishStat Jan 4 '17 at 15:42
• I recommend Brockwell and Davis "Time Series: Theory and Methods 2nd Edition" Springer 1991. – Michael R. Chernick Jan 4 '17 at 20:41

I would recommed the following books:

I hope it helps you. Best of luck!

• (+1) I've found the first book you listed there to be very useful. – Macro Jan 3 '12 at 3:19
• Biostat, could you clarify WHY you would recommend those books, above others? – naught101 Mar 7 '12 at 0:01
• or you, @Macro, considering this is a community wiki? – naught101 Mar 27 '12 at 0:49
• very good books, but maybe something more easy to undestood is also there? – user1406647 Oct 18 '15 at 22:03
• if we go by Amazon reviews, neither of these books proved friendly, if at all, to beginners, let alone self-learning beginners. – stucash Apr 26 '18 at 10:40

Forecasting: principles and practice by Rob J Hyndman and George Athanasopoulos is available free online: http://otexts.com/fpp/

It's a good book in its own right; Hyndman's previous forecasting book with Makridakis and Wheelright is highly regarded, but this has the added advantage that you can see what you're getting for the price.

There are three books that I keep referring to always from an R programming and time series analysis perspective:

1. Time Series Analysis and Its Applications: With R Examples by Shumway and Stoffer
2. Time Series Analysis: With Applications in R by Cryer and Chan.
3. Introductory Time Series with R by Cowpertwait and Metcalfe

The first book by Shumway and Stoffer has an open source (abridged) version available online called EZgreen version.

If you are specifically looking into time series forecasting, I would recommend following books:

1. Forecasting Methods and Applications by Makridakis, Wheelwright and Hyndman. I keep referring to this book repeatedly, This is a classic, writing style is absolutely phenomenal.
2. An online successor to the above book with nice R examples is Forecasting Principles and Practice by Hyndman and Athana­sopou­los.
3. If you are looking at classic Box Jenkins modeling approach, I would recommend Time Series Analysis: Forecasting and Control by Box, Jenkins and Reinsel.
4. An exceptional treatment on transfer function modeling and forecasting is in Forecasting with Dynamic Regression Models by Pankratz. Again the writing style is absolutely great.
5. Another extremely useful if you in to applying forecasting to solve real world problems is Principles of Forecasting by Armstrong.

In my opinion, books 1, 4 and 5 are some of the best of the best books. Many like Forecasting Principles and Practice by Hyndman and Athana­sopou­los because it's open source and has R codes. It is no way closer to the breadth, the depth of coverage of forecasting methods and the writing style of it predecessor Makridakis et al.. Below are some contrasting features on why I like the Makridakis et al:

1. List of references: for instance in the Box Jenkins chapter Makridakis et al has ~31 references, Hyndman et al there is very little or no references in many chapters.
2. Breadth and Depth in coverage - Hyndman et al. mainly focus on Univariate methods especially developed by the first author, while Makridakis et. al focus not just on their own research but a wide variety of methods and application and also emphasis is on real world application and learning as opposed to being more academically focused.
3. Writing style - I really cant complain as both the books are exceptionally well written. However I personally lean towards Makridakis because it boils down complex concepts into reader friendly sections. There is a section on Dynamic regression or transfer functions, I have no where encountered such clear explanation on this "complex method". It takes extraordinary writing talent to help reader understand what Dynamic regression is in 15 pages and they succeed at it.
4. Makridakis et al is software/method agnostic and they list some useful software packages and compare and contrast them (although this is almost 20 years old) is still a very valuable for a practitioner.
5. Three dedicated chapters on how to apply forecasting in real world in Makridakis et al. which is big plus to have for a practitioner.

Forecasting is simply not running univariate methods like arima and exponential smoothing and producing outputs. It is much more than that, and especially strategic forecasting when you are looking into longer horizon. Principles of forecasting by Armstrong goes beyond the univariate extrapolation methods and is highly recommended for anyone who does real world forecasting especially strategic forecasting.

• Hi, as you seem to be very expert on the subject, I would love to have your opinion on the book "Time Series Analysis, Forecasting and control" of Box et. al. I am new to time series analysis and have a PhD in applied mathematics (but very little knowledge in statistics) and know some machine learning. Would you recommend it? Or should I really start with the Makridakis? – Surb May 7 at 22:38
• @Surb if you like applied view of time series analysis and forecasting I would recommend Makridakis et al. if you like to learn more on theoritical aspects of ARIMA then Box et al. would be good. – forecaster May 16 at 13:00
• Thanks a lot for your reply. I am indeed more interested by the theoretical side currently, but in the end I will probably get both :). – Surb May 16 at 21:37

It depends on how much math you want. For a less mathematically-intense treatment, Applied Econometric Time Series by Enders is well-regarded.

Part Four of Damodar Gujarati and Dawn Porter's Basic Econometrics (5th ed) contains five chapters on time-series econometrics - a very popular book! It contains lots of exercises, regression outputs, interpretations, and best of all, you can download the data from the book's website and replicate the results for yourself. Another good book is Stock and Watson's Introduction to Econometrics.

Starting with Hamilton was admirable, but I'd say read through both of the time-series sections in the two books that I just mentioned and then move on to something like Walter Enders' Applied Econometric Time Series or Terrence C Mill's The Modelling of Financial Time Series.

After this (and probably after some review of mathematical economics) then you should be able to sit down and read Hamilton comfortably.

Note: Box & Jenkins' 1970 classic Time series analysis: Forecasting and control is obviously more concentrated (i.e. narrower in content) than the "modern textbooks" that I mentioned, but I'd say that anyone who wants to get a real good understanding of time-series shouldn't leave this off their reading list.

In addition to the other text there are two books introductory books in Springer's Use R! series that cover time series:
Introductory Time Series with R and Applied Econometrics in R

There is also an advanced econometrics text in the series, Analysis of Integrated and Co-integrated Time Series with R.

I have not used these but have found several others in the series to be excellent.

There are some good, free, online resources:

1. The Little Book of R for Time Series, by Avril Coghlan (also available in print, reasonably cheap) - I haven't read through this all, but it looks like it's well written, has some good examples, and starts basically from scratch (ie. easy to get into).
2. Chapter 15, Statistics with R, by Vincent Zoonekynd - Decent intro, but probably slightly more advanced. I find that there's too much (poorly commented) code, and not enough explanation thereof.

If you find Hamilton too difficult then there is Introduction to Econometric Modeling Princeton Uni Press by Bent Nielsen and David Hendry. It focuses more on intuition and practical how-tos than deeper theory. So if you're on a time constraint then that would be a good approach.

I would still recommend to persevere with Time Series Analysis by Hamilton. It is very deep mathematically and the first four chapters will keep you going for a long time and serve as a very strong introduction to the topic. It also covers Granger non-causality and cointegration and if you decide to pursue this topic more deeply then it is in invaluable resource.

For a more intuitive treatment of cointegration, I would also recommend Cointegration, Causality, and Forecasting by Engle and White.

Finally for very advanced treatments, there is Soren Johansen's book "Likelihood-Based Inference in Cointegrated VARs" and of course David Hendry's "Dynamic Econometrics".

Among those two, I would think Hendry's is more big-picture oriented and Johansen is pretty hard-going on the math.

• Hirek, did you notice the first sentence of the question, where the poster explains that they're already using Hamilton and don't understand it ... and want something else? – Glen_b -Reinstate Monica Mar 14 '15 at 14:35
• Ha totally overlooked that sorry @Glen_b – Hirek Mar 14 '15 at 16:44

Time Series Analysis: Univariate and Multivariate Methods by William Wei and David P. Reilly - is a very good book on time series and quite inexepnsive. There is am updated version but at a much higher price. It does not include R examples. It explicitely includes a great discussion/presentation of Intervention Detection procedures which are ignored in simplified solutions/introductory textbooks.

• The book gets good reviews, no complaints there. But I wonder whether you might have some relationship to one of the authors. Is that true? – whuber Jan 6 '17 at 16:38
• Yes that is true. I was one of the two authors . – IrishStat Jan 6 '17 at 17:03

There's the NBER Summer Institute "What's New in Time Series Econometrics" (not sure whether this material is gated or not). There are videos with accompanying slides. The lectures are given by a pair of professors (Stock and Watson) who are known for their popular undergraduate econometrics textbook.

In my opinion, you really can't beat Forecasting: principles and practice. It's written by CV's own Rob Hyndman and George Athana­sopou­los, it's available for free online, and it's got tons of example code in R, making use of the excellent forecast package.

• Zach, You might find this interesting. bit.ly/1Be6y4c – Tom Reilly Jun 18 '15 at 13:48
• @TomReilly Whatever the issues with any particular model, I'd still recommend the R language in general and the forecast package in particular to anyone looking to learn time series analysis. You really can't beat free, especially if your goal is education. – Zach Jun 18 '15 at 14:36
• Free purchase is one thing BUT if it contains trivial/uncomplicated/insufficient procedures to deal with non-simulated data you may have to subsequently/ultimately pay a price. – IrishStat Jun 19 '15 at 14:30
• @IrishStat Every dataset in FPP is non-simulated. Seems like great data to learn on... – Zach Jun 19 '15 at 15:56
• As long as you check to see if the residuals from the proposed model are free of structure otherwise the model may be insufficient as that structure should/can be transferred to the model . Even better training sets can be found in the AUTOBOX demo from 10 plus textbooks. Can't beat the price as it costs nothing , You should like it.. – IrishStat Jun 19 '15 at 16:00

If you use Stata, Introduction to Time Series Using Stata by Sean Becketti is a solid gentle introduction, with many examples and an emphasis on intuition over theory. I think this book would complement Ender rather well.

The book opens with an intro to Stata language, followed by a quick review of regression and hypothesis testing.

The time series part starts with moving-average and Holt–Winters techniques to smooth and forecast the data. The next section focuses on using these for techniques forecasting. These methods are often neglected, but they work rather well for automated forecasting and are easy to explain. Becketti explains when they will work and when they won't.

The next chapters cover single-equation time-series models like autocorrelated disturbances, ARIMA, and ARCH/GARCH modeling.

In the end, Becketti discusses multiple-equation models, particularly VARs and VECs, and non-stationary time series.

There are a few books that might be useful. If you are mathematically challenged you might want to start with two SAGE books by Mcdowall, Mcleary, Meidinger and Hay called "Interrupted Time Series Analysis" 1980 OR "Applied Time Series Analysis" by Richard McLeary. As you learn more about time series and decide that you you want more than prose and that you are willing to suffer through some math the Wei text published by Addison-Wessley entitled "Time Series Analysis" would be an excellent choice. In terms of web-based educational material, I have written a lot of useful material which can be viewed at http://www.autobox.com/cms/index.php/afs-university/intro-to-forecasting entitled "Introduction to Forecasting".

HILL GRIFFITHS LIM 2011 "Principles of Econometrics" 4E Wiley
(1) Very easy to follow. Topics are well presented. Even though I did not take any econometric course in my life, I easily grasped introductory econometrics with the book.

(2) There are supplemantary books to understand HILL's book:
a. Using EViews for Principles of Econometrics
b. Using Excel for Principles of Econometrics
c. Using Gretl for Principles of Econometrics
d. Using Stata for Principles of Econometrics