1.2 The Nature of Time Series Data 3
and related time domain models are studied in Chapter 3, with the empha-
sis on classical, statistical, univariate linear regression. Special topics on time
domain analysis are covered in Chapter 5; these topics include modern treat-
ments of, for example, time series with long memory and GARCH models
for the analysis of volatility. The state-space model, Kalman filtering and
smoothing, and related topics are developed in Chapter 6.
Conversely, the frequency domain approach assumes the primary charac-
teristics of interest in time series analyses relate to periodic or systematic
sinusoidal variations found naturally in most data. These periodic variations
are often caused by biological, physical, or environmental phenomena of inter-
est. A series of periodic shocks may influence certain areas of the brain; wind
may affect vibrations on an airplane wing; sea surface temperatures caused
by El Ni˜no oscillations may affect the number of fish in the ocean. The study
of periodicity extends to economics and social sciences, where one may be
interested in yearly periodicities in such series as monthly unemployment or
monthly birth rates.
In spectral analysis, the partition of the various kinds of periodic variation
in a time series is accomplished by evaluating separately the variance associ-
ated with each periodicity of interest. This variance profile over frequency is
called the power spectrum. In our view, no schism divides time domain and
frequency domain methodology, although cliques are often formed that advo-
cate primarily one or the other of the approaches to analyzing data. In many
cases, the two approaches may produce similar answers for long series, but
the comparative performance over short samples is better done in the time
domain. In some cases, the frequency domain formulation simply provides a
convenient means for carrying out what is conceptually a time domain calcu-
lation. Hopefully, this book will demonstrate that the best path to analyzing
many data sets is to use the two approaches in a complementary fashion. Ex-
positions emphasizing primarily the frequency domain approach can be found
in Bloomfield (1976, 2000), Priestley (1981), or Jenkins and Watts (1968).
On a more advanced level, Hannan (1970), Brillinger (1981, 2001), Brockwell
and Davis (1991), and Fuller (1996) are available as theoretical sources. Our
coverage of the frequency domain is given in Chapters 4 and 7.
The objective of this book is to provide a unified and reasonably complete
exposition of statistical methods used in time series analysis, giving serious
consideration to both the time and frequency domain approaches. Because a
myriad of possible methods for analyzing any particular experimental series
can exist, we have integrated real data from a number of subject fields into
the exposition and have suggested methods for analyzing these data.
1.2 The Nature of Time Series Data
Some of the problems and questions of interest to the prospective time se-
ries analyst can best be exposed by considering real experimental data taken