Time Series Analysis and

Its Applications
R.H. Shumway and D.S. Stoffer

Contents

Chapter 1: Characteristics of Time Series

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 The Nature of Time Series Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.3 Time Series From Statistical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.4 Measures of Dependence: Auto and Cross Correlation . . . . . . . . . . . . . . 15

1.5 Stationary Time Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.6 Estimation of Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

1.7 Exploratory Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1.8 Classical Regression and Smoothing in the Time Series Context. . . . . 39

1.9 Vector Valued and Multidimensional Series . . . . . . . . . . . . . . . . . . . . . . . . . 51

T1.10 Convergence Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

T1.11 Central Limit Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

T1.12 The Mean and Autocorrelation Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . 69

Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
 iv Contents

Chapter 2: Time Series Regression and ARIMA Models

2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.2 Autoregressive Moving Average Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.3 Homogeneous Difference Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.4 Autocorrelation and Partial Autocorrelation Functions . . . . . . . . . . . . 108
2.5 Forecasting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
2.6 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
2.7 Integrated Models for Nonstationary Data . . . . . . . . . . . . . . . . . . . . . . . . . 142
2.9 Building ARIMA Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
2.9 Multiplicative Seasonal ARIMA Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
2.10 Long Memory ARMA and Fractional Differencing . . . . . . . . . . . . . . . . . 166
2.11 Threshold Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.12 Regression With Autocorrelated Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2.13 Lagged Regression: Transfer Function Modeling . . . . . . . . . . . . . . . . . . . 178
2.14 ARCH Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
T2.15 Hilbert Spaces and the Projection Theorem . . . . . . . . . . . . . . . . . . . . . . . 190
T2.16 Causal Conditions for ARMA Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
T2.17 Large Sample Distribution of AR Estimators . . . . . . . . . . . . . . . . . . . . . . 197
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

Chapter 3: Spectral Analysis and Filtering

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

3.2 Cyclical Behavior and Periodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.3 Power Spectrum and Cross Spectrum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
3.4 Linear Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
3.5 Discrete Fourier Transform, Periodogram . . . . . . . . . . . . . . . . . . . . . . . . . 234
3.6 Nonparametric Spectral Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
3.7 Parametric Spectral Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
 Contents v

3.8 Lagged Regression Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

3.9 Signal Extraction and Optimum Filtering . . . . . . . . . . . . . . . . . . . . . . . . . 262
3.10 Spectral Analysis of Multidimensional Series . . . . . . . . . . . . . . . . . . . . . . 267
T3.11 Spectral Representation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
T3.12 Large Sample Distribution of Discrete Fourier Transform . . . . . . . . . . 274
T3.13 Complex Multivariate Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . 283
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288

Chapter 4: State Space and Multivariate ARMAX Models

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

4.2 Filtering, Smoothing and Forecasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
4.3 Maximum Likelihood Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
4.4 Missing Data Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
4.5 Structural Models: Signal Extraction and Forecasting . . . . . . . . . . . . . 331
4.6 ARMAX Models in State Space Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
4.7 Bootstrapping State-Space Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
4.8 Dynamic Linear Models With Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
4.9 Nonlinear and Nonnormal State-Space Models Using
Monte Carlo Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.10 Stochastic Volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
4.11 State Space and ARMAX Models for Longitudinal
Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
4.12 Further Aspects of Multivariate ARMA and ARMAX Models . . . . . 381
Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402

Chapter 5: Statistical Methods in the Frequency Domain

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410

5.2 Spectral Matrices and Likelihood Functions . . . . . . . . . . . . . . . . . . . . . . . 414
5.3 Regression for Jointly Stationary Series. . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
vi Contents

5.4 Regression with Deterministic Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423

5.5 Random Coefficient Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432

5.6 Analysis of Designed Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436

5.7 Discrimination and Cluster Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447

5.8 Principal Components, Canonical and Factor Analysis . . . . . . . . . . . . . 461

5.9 The Spectral Envelope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479

5.10 Dynamic Fourier Analysis and Wavelets . . . . . . . . . . . . . . . . . . . . . . . . . . . 502

Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
Preface

The goals of this book are to develop an appreciation for the richness and
versatility of modern time series analysis as a tool for analyzing data, and still
maintain a commitment to theoretical integrity, as exemplified by the seminal
works of Brillinger (1981) and Hannan (1970) and the texts by Brockwell and
Davis (1991) and Fuller (1995). The advent of more powerful computing, es-
pecially in the last three years, has provided both real data and new software
that can take one considerably beyond the fitting of simple time domain mod-
els, such as have been elegantly described in the landmark work of Box and
Jenkins (see Box et al, 1994). This book is designed to be useful as a text
for courses in time series on several different levels and as a reference work
for practitioners facing the analysis of time-correlated data in the physical,
biological, and social sciences.
We believe the book will be useful as a text at both the undergraduate and
graduate levels. An undergraduate course can be accessible to students with a
background in regression analysis and might include Sections 1.1-1.8, 2.1-2.9,
and 3.1-3.8. Similar courses have been taught at the University of California
(Berkeley and Davis) in the past using the earlier book on applied time series
analysis by Shumway (1988). Such a course is taken by undergraduate students
in mathematics, economics, and statistics and attracts graduate students from
the agricultural, biological, and environmental sciences. At the masters’ degree
level, it can be useful to students in mathematics, environmental science, eco-
nomics, statistics, and engineering by adding Sections 1.9, 2.10-2.14, 3.9, 3.10,
4.1-4.5, to those proposed above. Finally, a two-semester upper-level graduate
course for mathematics, statistics and engineering graduate students can be
crafted by adding selected theoretical sections from the last sections of Chap-
ters 1, 2, and 3 for mathematics and statistics students and some advanced
applications from Chapters 4 and 5. For the upper-level graduate course, we
should mention that we are striving for a less rigorous level of coverage than
that which is attained by Brockwell and Davis (1991), the classic entry at this
level.
A useful feature of the presentation is the inclusion of data illustrating
the richness of potential applications to medicine and in the biological, phys-
ical, and social sciences. We include data analysis in both the text examples
and in the problem sets. All data sets are posted on the World Wide Web
viii Preface

at the following URLs: http://www.stat.ucdavis.edu/~shumway/tsa.html

and http://www.stat.pitt.edu/~stoffer/tsa.html, making them easily
accessible to students and general researchers. In addition, an exploratory
data analysis program written by McQuarrie and Shumway (1994) can be
downloaded (as Freeware) from these websites to provide easy access to all of
the techniques required for courses through the masters’ level.
Advances in modern computing have made multivariate techniques in the
time and frequency domain, anticipated by the theoretical developments in
Brillinger (1981) and Hannan (1970), routinely accessible using higher level
languages, such as MATLAB and S-PLUS. Extremely large data sets driven by
periodic phenomena, such as the functional magnetic resonance imaging series
or the earthquake and explosion data, can now be handled using extensions
to time series of classical methods, like multivariate regression, analysis of
variance, principal components, factor analysis, and discriminant or cluster
analysis. Chapters 4 and 5 illustrate some of the immense potential that
methods have for analyzing high-dimensional data sets.
The many practical data sets are the results of collaborations with research
workers in the medical, physical, and biological sciences. Some deserve special
mention as a result of the pervasive use we have made of them in the text.
The predominance of applications in seismology and geophysics is joint work
of the first author with Dr. Robert R. Blandford of the Center for Monitoring
Research and Dr. Zoltan Der of Ensco, Inc. We have also made extensive use of
the El Niño and Recruitment series contributed by Dr. Roy Mendelssohn of the
National Marine Fisheries. In addition, Professor Nancy Day of the University
of Pittsburgh provided the data used in Chapter 4 in a longitudinal analysis
of the effects of prenatal smoking on growth, as well as some of the categorical
sleep-state data posted on the World Wide Web. A large magnetic imaging
data set that was developed during joint research on pain perception with Dr.
Elizabeth Disbrow of the University of San Francisco Medical Center forms the
basis for illustrating a number of multivariate techniques in Chapter 5. We
are especially indebted to Professor Allan D.R. McQuarrie of the University of
North Dakota, who incorporated subroutines in Shumway (1988) into ASTSA
for Windows.
Finally, we are grateful to John Kimmel, Executive Editor, Statistics, for
his patience, enthusiasm, and encouragement in guiding the preparation and
production of this book. Three anonymous reviewers made numerous helpful
comments, and Dr. Rahman Azari and Dr. Mitchell Watnik of the University
of California, Davis, Division of Statistics, read portions of the draft. Any
remaining errors are solely our responsibility.

Robert H. Shumway
Davis, CA
David S. Stoffer
Pittsburgh, PA
August, 1999