(eBook PDF) The Analysis of Time Series: An

Introduction with R 7th Edition pdf download

https://ebooksecure.com/product/ebook-pdf-the-analysis-of-time-
series-an-introduction-with-r-7th-edition/

Download more ebook from https://ebooksecure.com

We believe these products will be a great fit for you. Click
the link to download now, or visit ebooksecure.com
to discover even more!

(eBook PDF) Time Series: A Data Analysis Approach Using

R

http://ebooksecure.com/product/ebook-pdf-time-series-a-data-
analysis-approach-using-r/

Analysis with an introduction to proof. Fifth Edition,

Pearson New International Edition Steven R. Lay - eBook
PDF

https://ebooksecure.com/download/analysis-with-an-introduction-
to-proof-ebook-pdf/

(eBook PDF) An Introduction to Analysis 4th Edition by

William R. Wade

http://ebooksecure.com/product/ebook-pdf-an-introduction-to-
analysis-4th-edition-by-william-r-wade/

Understanding Price Action: practical analysis of the

5-minute time frame

http://ebooksecure.com/product/understanding-price-action-
practical-analysis-of-the-5-minute-time-frame/
Empowerment Series: An Introduction to the Profession
of Social Work - eBook PDF

https://ebooksecure.com/download/empowerment-series-an-
introduction-to-the-profession-of-social-work-ebook-pdf/

(eBook PDF) Biostatistics with R An Introduction to

Statistics Through Biological Data

http://ebooksecure.com/product/ebook-pdf-biostatistics-with-r-an-
introduction-to-statistics-through-biological-data/

(eBook PDF) An Introduction to Statistical Methods &

Data Analysis 7th

http://ebooksecure.com/product/ebook-pdf-an-introduction-to-
statistical-methods-data-analysis-7th/

(eBook PDF) Empowerment Series: An Introduction to the

Profession of Social Work 5th Edition

http://ebooksecure.com/product/ebook-pdf-empowerment-series-an-
introduction-to-the-profession-of-social-work-5th-edition/

(eBook PDF) Empowerment Series An Introduction to the

Profession of Social Work 6th Edition

http://ebooksecure.com/product/ebook-pdf-empowerment-series-an-
introduction-to-the-profession-of-social-work-6th-edition/
 Contents

Preface to the Seventh Edition xiii

Abbreviations and Notation xv

1 Introduction 1
1.1 Some Representative Time Series 1
1.2 Terminology 8
1.3 Objectives of Time Series Analysis 9
1.4 Approaches to Time Series Analysis 11
1.5 Review of Books on Time Series 12

2 Basic Descriptive Techniques 15

2.1 Types of Variation 15
2.2 Stationary Time Series 17
2.3 The Time Plot 17
2.4 Transformations 18
2.5 Analysing Series that Contain a Trend and No Seasonal
Variation 19
2.5.1 Curve fitting 20
2.5.2 Filtering 21
2.5.3 Differencing 25
2.5.4 Other approaches 25
2.6 Analysing Series that Contain a Trend and Seasonal Variation 25
2.7 Autocorrelation and the Correlogram 28
2.7.1 The correlogram 30
2.7.2 Interpreting the correlogram 31
2.8 Other Tests of Randomness 36
2.9 Handling Real Data 37

3 Some Linear Time Series Models 41

3.1 Stochastic Processes and Their Properties 41
3.2 Stationary Processes 42
3.3 Properties of the Autocorrelation Function 44
3.4 Purely Random Processes 45
3.5 Random Walks 47
3.6 Moving Average Processes 47

vii
viii CONTENTS
3.6.1 Stationarity and autocorrelation function of an MA
process 48
3.6.2 Invertibility of an MA process 49
3.7 Autoregressive Processes 52
3.7.1 First-order process 53
3.7.2 General-order process 54
3.8 Mixed ARMA Models 59
3.8.1 Stationarity and invertibility conditions 60
3.8.2 Yule-Walker equations and autocorrelations 60
3.8.3 AR and MA representations 62
3.9 Integrated ARMA (or ARIMA) Models 63
3.10 Fractional Differencing and Long-Memory Models 64
3.11 The General Linear Process 69
3.12 Continuous Processes 69
3.13 The Wold Decomposition Theorem 70

4 Fitting Time Series Models in the Time Domain 77

4.1 Estimating Autocovariance and Autocorrelation Functions 77
4.1.1 Using the correlogram in modelling 80
4.1.2 Estimating the mean 80
4.1.3 Ergodicity 81
4.2 Fitting an Autoregressive Process 81
4.2.1 Estimating parameters of an AR process 82
4.2.2 Determining the order of an AR process 84
4.3 Fitting a Moving Average Process 88
4.3.1 Estimating parameters of an MA process 88
4.3.2 Determining the order of an MA process 90
4.4 Estimating Parameters of an ARMA Model 94
4.5 Model Identification Tools 97
4.6 Testing for Unit Roots 99
4.7 Estimating Parameters of an ARIMA Model 102
4.8 Box–Jenkins Seasonal ARIMA Models 103
4.9 Residual Analysis 107
4.10 General Remarks on Model Building 110

5 Forecasting 115
5.1 Introduction 115
5.2 Extrapolation and Exponential Smoothing 117
5.2.1 Extrapolation of trend curves 118
5.2.2 Simple exponential smoothing 118
5.2.3 The Holt and Holt–Winters forecasting procedures 120
5.3 The Box–Jenkins Methodology 123
5.3.1 The Box-Jenkins procedures 123
5.3.2 Other methods 127
5.3.3 Prediction intervals 128
CONTENTS ix
5.4 Multivariate Procedures 135
5.4.1 Multiple regression 135
5.4.2 Econometric models 137
5.4.3 Other multivariate models 138
5.5 Comparative Review of Forecasting Procedures 138
5.5.1 Forecasting competitions 139
5.5.2 Choosing a non-automatic method 141
5.5.3 A strategy for non-automatic univariate forecasting 143
5.5.4 Summary 144
5.6 Prediction Theory 145

6 Stationary Processes in the Frequency Domain 149

6.1 Introduction 149
6.2 The Spectral Distribution Function 149
6.3 The Spectral Density Function 154
6.4 The Spectrum of a Continuous Process 157
6.5 Derivation of Selected Spectra 158

7 Spectral Analysis 167

7.1 Fourier Analysis 167
7.2 A Simple Sinusoidal Model 168
7.3 Periodogram Analysis 172
7.3.1 The relationship between the periodogram and the
autocovariance function 175
7.3.2 Properties of the periodogram 175
7.4 Some Consistent Estimation Procedures 177
7.4.1 Transforming the truncated autocovariance function 177
7.4.2 Hanning 179
7.4.3 Hamming 180
7.4.4 Smoothing the periodogram 180
7.4.5 The fast Fourier transform (FFT) 183
7.5 Confidence Intervals for the Spectrum 185
7.6 Comparison of Different Estimation Procedures 186
7.7 Analysing a Continuous Time Series 191
7.8 Examples and Discussion 193

8 Bivariate Processes 199

8.1 Cross-Covariance and Cross-Correlation 199
8.1.1 Examples 201
8.1.2 Estimation 202
8.1.3 Interpretation 203
8.2 The Cross-Spectrum 204
8.2.1 Examples 206
8.2.2 Estimation 209
8.2.3 Interpretation 211
x CONTENTS
9 Linear Systems 217
9.1 Introduction 217
9.2 Linear Systems in the Time Domain 219
9.2.1 Some types of linear systems 219
9.2.2 The impulse response function: An explanation 221
9.2.3 The step response function 222
9.3 Linear Systems in the Frequency Domain 223
9.3.1 The frequency response function 223
9.3.2 Gain and phase diagrams 227
9.3.3 Some examples 229
9.3.4 General relation between input and output 231
9.3.5 Linear systems in series 236
9.3.6 Design of filters 237
9.4 Identification of Linear Systems 238
9.4.1 Estimating the frequency response function 240
9.4.2 The Box–Jenkins approach 243
9.4.3 Systems involving feedback 247

10 State-Space Models and the Kalman Filter 253

10.1 State-Space Models 253
10.1.1 The random walk plus noise model 256
10.1.2 The linear growth model 256
10.1.3 The basic structural model 257
10.1.4 State-space representation of an AR(2) process 258
10.1.5 Bayesian forecasting 259
10.1.6 A regression model with time-varying coefficients 260
10.1.7 Model building 260
10.2 The Kalman Filter 261

11 Non-Linear Models 267

11.1 Introduction 267
11.1.1 Why non-linearity? 267
11.1.2 What is a linear model? 270
11.1.3 What is a non-linear model? 271
11.1.4 What is white noise? 272
11.2 Non-Linear Autoregressive Processes 273
11.3 Threshold Autoregressive Models 274
11.4 Smooth Transition Autoregressive Models 280
11.5 Bilinear Models 284
11.6 Regime-Switching Models 285
11.7 Neural Networks 290
11.8 Chaos 296
11.9 Concluding Remarks 300
11.10 Bibliography 301
CONTENTS xi
12 Volatility Models 303
12.1 Structure of a Model for Asset Returns 303
12.2 Historic Volatility 305
12.3 Autoregressive Conditional Heteroskedastic (ARCH) Models 306
12.4 Generalized ARCH Models 311
12.5 The ARMA-GARCH Models 315
12.6 Other ARCH-Type Models 318
12.6.1 The integrated GARCH model 319
12.6.2 The exponential GARCH model 320
12.7 Stochastic Volatility Models 320
12.8 Bibliography 321

13 Multivariate Time Series Modelling 323

13.1 Introduction 323
13.1.1 One equation or many? 324
13.1.2 The cross-correlation function 326
13.1.3 Initial data analysis 327
13.2 Single Equation Models 330
13.3 Vector Autoregressive Models 331
13.3.1 VAR(1) models 331
13.3.2 VAR(p) models 332
13.4 Vector ARMA Models 334
13.5 Fitting VAR and VARMA Models 335
13.6 Co-Integration 344
13.7 Multivariate Volatility Models 345
13.7.1 Exponentially weighted estimate 345
13.7.2 BEKK models 346
13.8 Bibliography 348

14 Some More Advanced Topics 351

14.1 Modelling Non-Stationary Time Series 351
14.2 Model Uncertainty 353
14.3 Control Theory 355
14.4 Miscellanea 356
14.4.1 Autoregressive spectrum estimation 357
14.4.2 Wavelets 357
14.4.3 ‘Crossing’ problems 358
14.4.4 Observations at unequal intervals, including missing
values 358
14.4.5 Outliers and robust methods 359
14.4.6 Repeated measurements 361
14.4.7 Aggregation of time series 361
14.4.8 Spatial and spatio-temporal series 362
14.4.9 Time series in finance 362
14.4.10 Discrete-valued time series 364
xii CONTENTS
Appendix A Fourier, Laplace, and z-Transforms 365

Appendix B Dirac Delta Function 369

Appendix C Covariance and Correlation 371

Answers to Exercises 373

References 381

Index 395
 Preface to the Seventh Edition

The first six editions of this book highlight basic concepts, models, and
methods in time series analysis, and have been used as a text for
undergraduate and graduate-level time series courses in many universities
during the past three decades. Although the previous editions successfully
introduce time series analysis in an accessible way, there is a small gap between
presenting time series theory and discussing its implementation, especially
given the fact that, recently, many statistical analyses can be easily carried
out with the aid of statistical software. I gradually realized this when I used the
sixth edition of the book for my time series courses during the past decade, and
felt that it would be much more convenient for both instructors and students
to have an introductory time series textbook that highlights not only basic
time series theory but the implementation of time series analysis as well.
Obviously, Chris Chatfield shared the same view with me. The concrete
idea of having this edition of the book arose in 2016 when Chris Chatfield
and Rob Calver, Executive Editor in Mathematics, Statistics, and Physics at
Taylor and Francis, asked me if I was interested in revising the sixth edition
of the book. By then, I had been teaching undergraduate and graduate-level
time series courses for over ten years, and had collected a set of examples on
real data analysis with R implementation, so I expressed my interest to both
Rob and Chris and started working on the new edition.
Similar to the sixth edition, this edition assumes knowledge of basic
probability theory and elementary statistical inference. As the sixth edition of
the book covers a broad range of topics at the introductory level, this edition
keeps most of the material from the sixth edition. However, several changes
are made in this edition. First, a new chapter (Chapter 12) and a new section
(Section 13.7) are added to introduce uni- and multi-variate volatility models
in finance, respectively. Necessary updates are also made in different chapters
and sections. Second, many examples and real data are added in this edition.
Specifically, I added examples of real data analysis in most chapters except for
Chapters 9 and 10. Third, all examples in the book are implemented with R,
and R codes for most examples are provided in the book so that the reader can
easily replicate the result. The data and scripts in the book are available at
http://www.ams.sunysb.edu/~xing/tsRbook/index.html.
I would like to thank Chris Chatfield for his invitation and authorization
for revising the book. I also thank all the students who took my time series
course for their interest in the subject and comments on the earlier draft

xiii
xiv PREFACE TO THE SEVENTH EDITION
of the book. Besides, I want to express my gratitude to my colleagues for
being supportive and helpful over the years. At last, I want to thank the
U.S. National Science Foundation for providing support for my research and
teaching during the past years. Any errors, omissions, or obscurities in this
edition are my responsibility.

Haipeng Xing
Department of Applied Mathematics and Statistics
State University of New York, Stony Brook
Stony Brook, NY 11794, U.S.A.
e-mail: [email protected]
 Abbreviations and Notation

AR Autoregressive
MA Moving average
ARMA Autoregressive moving average
ARIMA Autoregressive integrated moving average
SARIMA Seasonal ARIMA
TAR Threshold autoregressive
GARCH Generalized autoregressive conditionally heteroscedastic
SWN Strict white noise
WN White noise

MMSE Minimum mean square error

P.I. Prediction interval
FFT Fast Fourier transform

ac.f. Autocorrelation function

acv.f. Autocovariance function

N Sample size, or length of observed time series

x
bN (h) Forecast of xN +h made at time N
B Backward shift operator such that BXt = Xt−1
Bd B d Xt = Xt−d
∇ First differencing operator, (1 − B), such that
∇Xt = Xt − Xt−1
∇d ∇d = 1 − B d
∇rd ∇rd = (1 − B d )r
E Expectation or expected value
Var Variance
I Identity matrix – a square matrix with ones on the
diagonal and zeros otherwise
N (µ, σ 2 ) Normal distribution with mean µ and variance σ 2
χ2ν Chi-square distribution with ν degrees of freedom
{Zt } or {εt } Purely random process of independent random variables,
usually N (0, σ 2 )-distributed
AT or X T Transpose of a matrix A or vector X — vectors are
indicated by boldface, but not scalars or matrices

xv
 Chapter 1

Introduction

A time series is a collection of observations made sequentially through time.

Examples occur in a variety of fields, ranging from economics to engineering,
and methods of analysing time series constitute an important area of statistics.

1.1 Some Representative Time Series

We begin with some examples of the sort of time series that arise in practice.

Economic and financial time series

Many time series are routinely recorded in economics and finance. Examples
include share prices on successive days, export totals in successive months,
average incomes in successive months, company profits in successive years
and so on.
The classic Beveridge wheat price index series consists of the average wheat
price in nearly 50 places in various countries measured in successive years
from 1500 to 1869 (Beveridge, 1921). This series is of particular interest to
economics historians, and is available in many places (e.g. in the tseries
package of R). Figure 1.1 shows this series and some apparent cyclic behaviour
can be seen. The trend of the series will be studied in Section 2.5.2.
To plot the data using the R statistical package, you can load the data bev
in the tseries package and plot the time series (the > below are prompts):

> library(tseries) # load the library

> data(bev) # load the dataset
> plot(bev, xlab="Year", ylab="Wheat price index", xaxt="n")
> x.pos<-c(1500, 1560, 1620, 1680, 1740, 1800, 1869)
# define x-axis labels
> axis(1, x.pos, x.pos)

As an example of financial time series, Figure 1.2 shows the daily returns
(or percentage change) of the adjusted closing prices of the Standard & Poor’s
500 (S&P500) Index from January 4, 1995 to December 30, 2016. The data
shown in Figure 1.2 are typical of return data. The mean of the return series
seems to be stable with an average return of approximately zero, but the
volatility of data changes over time. This series will be analyzed in Chapter 12.

1
2 INTRODUCTION

300
Wheat price index

200
100
0

1500 1560 1620 1680 1740 1800 1869

Year

Figure 1.1 The Beveridge wheat price annual index series from 1500 to 1869.
10
5
Daily return

0
−5

1/4/1995 5/7/2001 9/19/2007 1/28/2014

Day

Figure 1.2 Daily returns of the adjusted closing prices of the S&P500 index from
January 4, 1995 to December 30, 2016.
SOME REPRESENTATIVE TIME SERIES 3
To reproduce Figure 1.2 in R, suppose you save the data as
sp500 ret 1995-2016.csv in the directory mydata. Then you can use the
following command to read the data and plot the time series.

> sp500<-read.csv("mydata/sp500_ret_1995-2016.csv")
> n<-nrow(sp500)
> x.pos<-c(seq(1,n,800),n)
> plot(sp500$Return, type="l", xlab="Day",
ylab="Daily return", xaxt="n")
> axis(1, x.pos, sp500$Date[x.pos])

Physical time series

Many types of time series occur in the physical sciences, particularly
in meteorology, marine science and geophysics. Examples are rainfall on
successive days, and air temperature measured in successive hours, days or
months. Figure 1.3 shows the average air temperature in Anchorage, Alaska
in the United States in successive months over a 16-year period. The series can
be downloaded from the U.S. National Centers for Environmental Information
(https://www.ncdc.noaa.gov/cag/). Seasonal fluctuations can be clearly
seen in the series.

Some mechanical recorders take measurements continuously and produce

a continuous trace rather than observations at discrete intervals of time.
For example, in some laboratories it is important to keep temperature and
humidity as constant as possible and so devices are installed to measure
these variables continuously. Action may be taken when the trace goes
outside pre-specified limits. Visual examination of the trace may be adequate
for many purposes, but, for more detailed analysis, it is customary to
convert the continuous trace to a series in discrete time by sampling the
trace at appropriate equal intervals of time. The resulting analysis is more
straightforward and can readily be handled by standard time series software.

Marketing time series

The analysis of time series arising in marketing is an important problem in
commerce. Observed variables could include sales figures in successive weeks or
months, monetary receipts, advertising costs and so on. As an example, Figure
1.4 shows the domestic sales of Australian fortified wine by winemakers in
successive quarters over a 30-year period, which are available at the Australian
Bureau of Statistics (http://www.abs.gov.au/AUSSTATS/). This series will
be analysed in Sections 4.8 and 4.9. Note the trend and seasonal variation
which is typical of sales data. It is often important to forecast future sales so
as to plan production. It may also be of interest to examine the relationship
between sales and other time series such as advertising expenditure.
4 INTRODUCTION

10
Average temperature

0
−10
−20

01/2001 01/2005 01/2009 01/2013 12/2016

Month

Figure 1.3 Monthly average air temperature (deg C) in Anchorage, Alaska, the
United States, in successive months from 2001 to 2016.
10000
Sales in thousand liters

8000
6000
4000

Mar−1985 Mar−1990 Mar−1995 Mar−2000 Mar−2005 Mar−2010 Jun−2014

Month

Figure 1.4 Domestic sales (unit: thousand liters) of Australian fortified wine by
winemakers in successive quarters from March 1985 to June 2014.
SOME REPRESENTATIVE TIME SERIES 5

3.0e+08
Population
2.4e+08
1.8e+08

1960 1967 1974 1981 1988 1995 2002 2009 2015

Year
24
22
20
Birth rates
18
16
14
12

1960 1967 1974 1981 1988 1995 2002 2009 2015

Year

Figure 1.5 Total population and birth rate (per 1,000 people) for the United States
from 1965 to 2015.

Demographic time series

Various time series occur in the study of population change. Examples include
the total population of Canada measured annually, and monthly birth totals in
England. Figure 1.5 shows the total population and crude birth rate (per 1,000
people) for the United States from 1965 to 2015. The data are available at the
U.S. Federal Reserve Bank of St. Louis (https://fred.stlouisfed.org/).
Demographers want to predict changes in population for as long as 10 or 20
years into the future, and are helped by the slowly changing structure of a
human population. Standard time series methods can be applied to study this
problem.
To reproduce Figure 1.5 in R, you can use the following command to read
the data and plot the time series.
> pop<-read.csv("mydata/US_pop_birthrate.csv", header=T)
> x.pos<-c(seq(1, 56, 7), 56)
> x.label<-c(seq(1960, 2009, by=7), 2015)

> par(mfrow=c(2,1), mar=c(3,4,3,4))

> plot(pop[,2], type="l", xlab="", ylab="", xaxt="n")
> points(pop[,2])
> axis(1, x.pos, x.label, cex.axis=1.2)
> title(xlab="Year", ylab="Population", line=2, cex.lab=1.2)
6 INTRODUCTION

Process variable

Target value

Time

Figure 1.6 A process control chart.

> plot(pop[,3], type="l", xlab="", ylab="", xaxt="n")

> points(pop[,3])
> axis(1, x.pos, x.label, cex.axis=1.2)
> title(xlab="Year", ylab="Birth rates", line=2, cex.lab=1.2)

Process control data

In process control, a problem is to detect changes in the performance of a
manufacturing process by measuring a variable, which shows the quality of
the process. These measurements can be plotted against time as in Figure 1.6.
When the measurements stray too far from some target value, appropriate
corrective action should be taken to control the process. Special techniques
have been developed for this type of time series problems, and the reader is
referred to a book on statistical quality control (e.g. Montgomery, 1996).

Binary processes
A special type of time series arises when observations can take one of only two
values, usually denoted by 0 and 1 (see Figure 1.7). For example, in computer
science, the position of a switch, either ‘on’ or ‘off’, could be recorded as one
or zero, respectively. Time series of this type, called binary processes, occur
in many situations, including the study of communication theory. A problem
here is to predict when the process will take a different value. One way to
solve this problem is to use regime-switching models, which will be discussed
in Chapter 11 (Section 11.6).
SOME REPRESENTATIVE TIME SERIES 7

0
Time

Figure 1.7 A realization of a binary process.

Point processes
A completely different type of time series occurs when we consider a series of
events occurring ‘randomly’ through time. For example, we could record the
dates of major railway disasters. A series of events of this type is usually
called a point process. As an example, Figure 1.8 shows the intraday
transaction data of the International Business Machines Corporation (IBM)
from 9:35:00 to 9:38:00 on January 4, 2010. When a trade event occurs, the
corresponding trading price and trading volume are observed. However, trades
do not occur equally spaced in time; hence time intervals between trades (or
trade durations) are considered as random variables. For observations of this
type, we are interested in such quantities as the distribution of the number
of events occurring in a given time period and distribution of time intervals
between events. Methods of analysing point process data are generally very
different from those used for analysing standard time series data and the
reader is referred, for example, to Cox and Isham (1980).
To reproduce Figure 1.8 in R, you can use the following command to read
the data and plot the time series.
> ibm<-read.table("mydata/taq_trade_ibm_100104.txt",
header=T, sep="\t")
> ibm.new<-ibm[,c(1,2,7)]
> ibm[,2]<-as.numeric(as.character(ibm[,2]))

> ### take 9:35:00-9:37:59am trading record

> data<-ibm.new[1458:2371,]
> newtime<-rep(0, nrow(data))
> for (i in 1:nrow(data)){
min<-as.numeric(substr(as.character(data$TIME[i]),3,4))
sec<-as.numeric(substr(as.character(data$TIME[i]),6,7))
8 INTRODUCTION

132.0
131.8
Price
131.6

9:35:00 9:35:31 9:36:00 9:36:30 9:37:00 9:37:30 9:38:00

Time
2000
Volume
500 1000
0

9:35:00 9:35:31 9:36:00 9:36:30 9:37:00 9:37:30 9:38:00

Time

Figure 1.8 Transaction prices and volumes of IBM stocks from 9:35:00 to 9:38:00
on January 4, 2010.

newtime[i]<- (min-30)*60+sec
}

> x.label<-c("9:35:00", "9:35:31", "9:36:00", "9:36:30",

"9:37:00", "9:37:30", "9:38:00")
> x.pos<-c(1, 139, 249, 485, 619, 776, 914)

> par(mfrow=c(2,1), mar=c(2,4,2,4))

> plot(newtime, data[,2],xlab="",ylab="",xaxt="n",type="h")
> axis(1, newtime[x.pos], x.label, cex.axis=1.2)
> title(xlab="Time", ylab="Price", line=2, cex.lab=1.2)
> plot(newtime, data[,3],xlab="",ylab="",xaxt="n",type="h")
> axis(1, newtime[x.pos], x.label, cex.axis=1.2)
> title(xlab="Time", ylab="Volume", line=2, cex.lab=1.2)

1.2 Terminology
A time series is said to be continuous when observations are made
continuously through time as in Figure 1.7. The adjective ‘continuous’ is
used for series of this type even when the measured variable can only take
a discrete set of values, as in Figure 1.7. A time series is said to be discrete
when observations are taken only at specific times, usually equally spaced.
Another Random Scribd Document
with Unrelated Content
 GEORG FRIDERIC
HANDEL
(Born at Halle, February 23, 1685; died at London, April 14, 1759)

“Mr. Georg Frideric Handel,” Mr. Runciman once wrote, “is by far the
most superb personage one meets, in the history of music. He
alone, of all the musicians, lived his life straight through in the grand
[29]
manner.” When Handel wrote “pomposo” on a page, he wrote not
idly. What magnificent simplicity in outlines!... For melodic lines of
such chaste and noble beauty, such Olympian authority, no one has
approached Handel. “Within that circle none durst walk but he.” His
nearest rival is the Chevalier Gluck.

And this giant of a man could express a tenderness known only to

him and Mozart, for Schubert, with all his melodic wealth and
sensitiveness, could fall at times into sentimentalism, and
Schumann’s intimate confessions were sometimes whispered. Handel
in his tenderness was always manly. No one has approached him in
his sublimely solemn moments! Few composers, if there is anyone,
have been able to produce such pathetic or sublime effects by
simple means, by a few chords even. He was one of the greatest
melodists. His fugal pages seldom seem labored; they are
distinguished by amazing vitality and spontaneity. In his slow
movements, his instrumental airs, there is a peculiar dignity, a
peculiar serenity, and a direct appeal that we find in no other
composer.

Would that we could hear more of Handel’s music! At present he is

known in this country as the composer of The Messiah, the 151
variations entitled The Harmonious Blacksmith, and the monstrous
perversion of a simple operatic air dignified, forsooth, by the title
“Handel’s Largo.”
 TWELVE CONCERTI GROSSI, FOR STRING
ORCHESTRA

No. 1, in G major
No. 2, in F major
No. 3, in E minor
No. 4, in A minor
No. 5, in D major
No. 6, in G minor
No. 7, in B flat major
No. 8, in C minor
No. 9, in F major
No. 10, in D minor
No. 11, in A major
No. 12, in B minor

Handel apparently took a peculiar pride in his Concerti Grossi. He

published them himself, and by subscription. They would probably
be more popular today if all conductors realized the fact that music
in Handel’s time was performed with varied and free inflections; that
his players undoubtedly employed many means of expression. As
German organists of forty years ago insisted that Bach’s preludes,
fugues, toccatas, should be played with full organ and rigidity of
tempo, although those who heard Bach play admired his skill in
registration, many conductors find in all of the allegros of Handel’s
concertos only a thunderous speech and allow little change in
tempo. In the performance of this old music, old but fresh, the two
essential qualities demanded by Handel’s music, suppleness of pace
and fluidity of expression, named by Volbach, are usually
disregarded. Unless there be elasticity in performance, hearers are
not to be blamed if they find the music formal, monotonous, dull.

The twelve concertos were composed within three weeks.

Kretzschmar has described them as impressionistic pictures,
probably without strict reference to the modern use of the word
[30]
“impressionistic.” They are not of equal worth. Romain Rolland
finds the seventh and three last mediocre. In the tenth he 152
discovers French influences and declares that the last allegro
might be an air for a music box. Yet the music at its best is
aristocratic and noble.

Handel’s twelve grand concertos for strings were composed between

September 29 and October 30, 1739. The London Daily Post of
October 29, 1739, said: “This day are published proposals for
printing by subscription, with His Majesty’s royal license and
protection, Twelve Grand Concertos, in Seven Parts, for four violins,
a tenor, a violoncello, with a thorough-bass for the harpsichord.
Composed by Mr. Handel. Price to subscribers, two guineas. Ready
to be delivered by April next. Subscriptions are taken by the author,
at his house in Brook Street, Hanover Square, and by Walsh.” In an
advertisement on November 22 the publisher added, “Two of the
above concertos will be performed this evening at the Theatre Royal,
Lincoln’s Inn.” The concertos were published on April 21, 1740. In an
advertisement a few days afterwards Walsh said, “These concertos
were performed at the Theatre Royal in Lincoln’s Inn Fields, and now
are played in most public places with the greatest applause.” Victor
Schoelcher made this comment in his Life of Handel: “This was the
case with all the works of Handel. They were so frequently
performed at contemporaneous concerts and benefits that they
seem, during his lifetime, to have quite become public property.
Moreover, he did nothing which the other theaters did not attempt to
imitate. In the little theater of the Haymarket, evening
entertainments were given in exact imitation of his ‘several
concertos for different instruments, with a variety of chosen airs of
the best master, and the famous Salve Regina of Hasse.’ The
handbills issued by the nobles at the King’s Theatre make mention
[31]
also of ‘several concertos for different instruments.’”

The year 1739, in which these concertos were composed, was the
year of the first performance of Handel’s Saul (January 16) and
Israel in Egypt (April 4)—both oratorios were composed in 1738—
also of the music to Dryden’s Ode for St. Cecilia’s Day (November
22).

Romain Rolland, discussing the form concerto grosso, which consists

essentially of a dialogue between a group of soloists, the concertino
(trio of two solo violins and solo bass with cembalo) and the 153
chorus of instruments, concerto grosso, believes that Handel
at Rome in 1708 was struck by Corelli’s works in this field, for
several of his concertos of Opus 3 are dated 1710, 1716, 1722.
Geminiani introduced the concerto into England—three volumes
appeared in 1732, 1735, 1748—and he was a friend of Handel.

It is stated that the word “concerto,” as applied to a piece for a solo

instrument with accompaniment, first appeared in a treatise by
Scipio Bargaglia (Venice, 1587); that Giuseppe Torelli, who died in
1708, was the first to suggest a larger number of instruments in a
concerto, and to give the name concerto grosso to this species of
composition. But Michelletti, seventeen years before, had published
his Sinfonie e concerti a quatro, and in 1698 his Concerti musicali,
while the word “concerto” occurs frequently in the musical
terminology of the seventeenth century. It was Torelli who,
determining the form of the grand solo for violin, opened the way to
Archangelo Corelli, the father of modern violinists, composers, or
virtuosos.
Romain Rolland insisted that the instrumental music of Handel has
the nature of a constant improvisation, music to be served piping
hot to an audience, and should preserve this character in
performance. “When you have studied with minute care each detail,
obtained from your orchestra an irreproachable precision, tonal
purity, and finish, you will have done nothing unless you have made
the face of the improvising genius rise from the work.”

154
 FRANZ JOSEF
HAYDN
(Born at Rohrau, Lower Austria, March 31, 1732; died at Vienna,
May 31, 1809)

Haydn has been sadly misunderstood by present followers of

tradition who have spoken of him as a man of the old school, while
Mozart was a forerunner of Beethoven. Thus they erred. Mozart
summed up the school of his day and wrote imperishable music.
There has been only one Mozart, and there is no probability of
another being born for generations to come; but Haydn was often
nearer in spirit to the young Beethoven. It is customary to speak
lightly of Haydn as an honest Austrian who wrote light-hearted
allegros, also minuets by which one is not reminded of a court with
noble dames smiling graciously on gallant cavaliers, but sees
peasants thumping the ground with heavy feet and uttering joyful
cries.

It is said carelessly that Haydn was a simple fellow who wrote at

ease many symphonies and quartets that, to quote Berlioz, recall
“the innocent joys of the fireside and the pot-au-feu.” But Haydn was
shrewd and observing—read his diary, kept in London—and if he was
plagued with a shrewish wife he found favor with other women.
Dear Mrs. Schroeter of London received letters from him breathing
love, not manly complimentary affection. And it is said of Haydn that
he was only sportive in his music, having a fondness for the
bassoon. But Haydn could express tenderness, regret, sorrow in his
music.
 155

LONDON SYMPHONIES
SYMPHONY NO. 104, IN D MAJOR (B. & H. NO. 2)

I. Adagio; allegro
II. Andante
III. Menuetto; trio
IV. Allegro spiritoso

Haydn’s symphony is ever fresh, spontaneous, yet contrapuntally

worked in a masterly manner. What a skillful employment of little
themes in themselves of slight significance save for their Blakelike
innocence and gayety! Yet in the introduction there is a deeper note,
for, contrary to current and easy belief, Haydn’s music is not all beer,
skittles, and dancing. There are even gloomy pages in some of his
quartets; tragic pages in his Seven Last Words, and the prelude to
The Creation, depicting chaos, is singularly contemporaneous.

Haydn composed twelve symphonies in England for Salomon. His

name began to be mentioned in England in 1765. Symphonies by
him were played in concerts given by J. C. Bach, Abel, and others in
the ’seventies. Lord Abingdon tried in 1783 to persuade Haydn to
take the direction of the Professional Concerts which had just been
founded. Gallini asked him his terms for an opera. Salomon, violinist,
conductor, manager, sent a music publisher, one Bland—an
auspicious name—to coax him to London, but Haydn was loath to
leave Prince Esterhazy. Prince Nicolaus died in 1790, and his
successor, Prince Anton, who did not care for music, dismissed the
orchestra at Esterház and kept only a brass band; but he added 400
gulden to the annual pension of 1,000 gulden bequeathed to Haydn
by Prince Nicolaus. Haydn then made Vienna his home. And one day,
when he was at work in his house, the “Hamberger” house in which
Beethoven also once lived, a man appeared, and said: “I am
Salomon from London, and come to fetch you with me. We 156
will agree on the job tomorrow.” Haydn was intensely
amused by the use of the word “job.” The contract for one season
was as follows: Haydn should receive three hundred pounds for an
opera written for the manager Gallini, £300 for six symphonies and
£200 for the copyright, £200 for twenty new compositions to be
produced in as many concerts under Haydn’s direction, £200 as
guarantee for a benefit concert, Salomon deposited 5,000 gulden
with the bankers, Fries & Company, as a pledge of good faith. Haydn
had 500 gulden ready for traveling expenses, and he borrowed 450
more from his prince. Haydn agreed to conduct the symphonies at
the piano.

Salomon about 1786 began to give concerts as a manager, in

addition to fiddling at concerts of others. He had established a series
of subscription concerts at the Hanover Square Rooms, London. He
thought of Haydn as a great drawing card. The violinist W. Cramer,
associated with the Professional Concerts, had also approached
Haydn, who would not leave his prince. The news of Prince
Esterhazy’s death reached Salomon, who then happened to be at
Bonn. He therefore hastened to Vienna.

The first of the Salomon-Haydn concerts was given March 11, 1791,
at the Hanover Square Rooms. Haydn, as was the custom, “presided
at the harpsichord”; Salomon stood as leader of the orchestra. The
symphony was in D major, No. 2, of the London list of twelve. The
adagio was repeated, an unusual occurrence, but the critics
preferred the first movement.
The orchestra was thus composed: twelve to sixteen violins, four
violas, three violoncellos, four double basses, flute, oboe, bassoon,
horns, trumpets, drums—in all about forty players.

Haydn and Salomon left Vienna on December 15, 1790, and arrived
at Calais by way of Munich and Bonn. They crossed the English
Channel on New Year’s Day, 1791. From Dover they traveled to
London by stage. The journey from Vienna took them seventeen
days. Haydn was received with great honor.

Haydn left London towards the end of June, 1792. Salomon invited
him again to write six new symphonies. Haydn arrived in London,
February 4, 1794, and did not leave England until August 15, 1795.
The orchestra at the opera concerts in the grand new concert hall of
the King’s Theatre was made up of sixty players. Haydn’s
engagement was again a profitable one. He made by concerts,
lessons, symphonies, etc., £1,200. He was honored in many 157
ways by the King, the Queen, and the nobility. He was
twenty-six times at Carlton House, where the Prince of Wales had a
concert room; and, after he had waited long for his pay, he sent a
bill from Vienna for 100 guineas, which Parliament promptly settled.
 LONDON SYMPHONIES
SYMPHONY NO. 94, IN G MAJOR, “SURPRISE” (B. & H. NO.

6)

I. Adagio cantabile e vivace assai

II. Andante
III. Menuetto
IV. Allegro di molto

This symphony, known as the “Surprise,” and in Germany as the

symphony “with the drumstroke,” is the third of the twelve Salomon
symphonies as arranged in the order of their appearance in the
catalogue of the Philharmonic Society (London).

Composed in 1791, this symphony was performed for the first time
on March 23, 1792, at the sixth Salomon concert in London. It
pleased immediately and greatly. The Oracle characterized the
second movement as one of Haydn’s happiest inventions, and
likened the “surprise”—which is occasioned by the sudden orchestral
crash in the andante—to a shepherdess, lulled by the sound of a
distant waterfall, awakened suddenly from sleep and frightened by
the unexpected discharge of a musket.

Griesinger in his Life of Haydn (1810) contradicts the story that

Haydn introduced these crashes to arouse the Englishwomen from
sleep. Haydn also contradicted it; he said it was his intention only to
surprise the audience by something new. “The first allegro of my
symphony was received with countless ‘Bravos,’ but enthusiasm rose
to its highest pitch after the andante with the drumstroke. ‘Ancora!
ancora!’ was cried out on all sides, and Pleyel himself complimented
me on my idea.” On the other hand, Gyrowetz, in his Autobiography,
page 59 (1848), said that he visited Haydn just after he had
composed the andante, and Haydn was so pleased with it 158
that he played it to him on the piano, and sure of his
success, said with a roguish laugh: “The women will cry out here!”
[32]
C. F. Pohl added a footnote, when he quoted this account of
Gyrowetz, and called attention to Haydn’s humorous borrowing of a
musical thought of Martini to embellish his setting of music to the
commandment, “Thou shalt not steal,” when he had occasion to put
music to the Ten Commandments. The Surprise symphony was long
known in London as “the favorite grand overture.”
 PARIS SYMPHONIES
SYMPHONY NO. 88, IN G MAJOR (B. & H. NO. 13)

I. Adagio; allegro
II. Largo
III. Menuetto; trio
IV. Finale; allegro con spirito

The Parisian orchestra, which Haydn undoubtedly had in mind, was a

large one—forty violins, twelve violoncellos, eight double basses—so
that the composer could be sure of strong contrasts in performance
by the string section. Fortunate composer—whose symphonies one
can, sitting back, enjoy without inquiring into psychological intention
or noting attempts at realism in musical seascapes and landscapes—
music not inspired by book or picture—just music; now pompous,
now merry, and in more serious moments, never too sad, but with a
constant feeling for tonal grace and beauty.

Haydn wrote a set of six symphonies for a society in Paris known as

the Concert de la loge olympique. They were ordered in 1784, when
Haydn was living at Esterház. Composed in the course of the years
1784-89, they are in C, G minor, E flat, B flat, D, A. No. 1, in C, has
been entitled the “Bear”; No. 2, in G minor, has been entitled 159
the “Hen”; and No. 4, in B flat, is known as the “Queen of
France.” This symphony is the first of a second set, of which five
were composed in 1787, 1788, 1790. If the sixth was written, it
cannot now be identified. This one in G major was written in 1787,
and is numbered 88 in the full and chronological listing of
Mandyczewski (given in Grove’s Dictionary).

I. The first movement opens with a short, slow introduction, adagio,

G major, 3-4 which consists for the most part of strong staccato
chords which alternate with softer passages. The main body of the
movement allegro, G major, begins with the first theme, a dainty
one, announced piano by the strings without double basses and
repeated forte by the full orchestra with a new counter figure in the
bass. A subsidiary theme is but little more than a melodic variation
of the first. So, too, the short conclusion theme—in oboes and
bassoon, then in the strings—is only a variation of the first. The free
fantasia is long for the period and is contrapuntally elaborate. There
is a short coda on the first theme.

II. Largo, D major, 3-4. A serious melody is sung by oboe and

violoncellos to an accompaniment of violas, double basses, bassoon,
and horn. The theme is repeated with a richer accompaniment;
while the first violins have a counter figure. After a transitional
passage the theme is repeated by a fuller orchestra, with the melody
in first violins and flute, then in the oboe and violoncello. The
development is carried along on the same lines. There is a very
short coda.

III. The Menuetto, allegretto, G major, 3-4, with trio, is in the regular
minuet form in its simplest manner.

IV. The finale, allegro con spirito, G major, 2-4, is a rondo on the
theme of a peasant country dance, and it is fully developed. Haydn
in his earlier symphonies adopted for the finale the form of his first
movement. Later he preferred the rondo form, with its couplets and
refrains, or repetitions of a short and frank chief theme. “In some
[33]
finales of his last symphonies,” says Brenet, “he gave freer reins
to his fancy, and modified with greater independence the form of his
first allegros; but his fancy, always prudent and moderate, is more
like the clear, precise arguments of a great orator than the headlong
inspiration of a poet. Moderation is one of the characteristics of
Haydn’s genius; moderation in the dimensions, in the sonority, in the
melodic shape; the liveliness of his melodic thought never 160
seems extravagant, its melancholy never induces sadness.”

The usual orchestration of Haydn’s symphonies (including those

listed above) consisted of one (or two) flutes, two oboes, two
bassoons, two horns, two trumpets, kettledrums, and strings. In his
last years (from 1791) he followed Mozart’s lead in introducing two
clarinets. The clarinets accordingly appear in the London symphony
in D major, described in this chapter.—EDITOR.

161
 PAUL
HINDEMITH
(Born at Hanau, on November 16, 1895)
 “KONZERTMUSIK” FOR STRING AND BRASS
INSTRUMENTS

There was a time in Germany when Hindemith was regarded as the

white-haired boy; the hope for the glorious future; greater even than
Schönberg. In England, they look on Hindemith coolly—an able and
fair-minded critic there has remarked: “The more one hears of the
later Hindemith, the more exasperating his work becomes. From
time to time some little theme is shown at first in sympathetic
fashion, then submitted to the most mechanical processes known to
music. Any pleasant jingle seems to mesmerize the composer, who
repeats it much as Bruckner repeats his themes—Hindemith abuses
the liberty shown to a modern.”

But Hindemith is not always mesmerized by a pleasant jingle.

Witness his oratorio, performed with great success. The title is
forbidding, The Unending, but the performance takes only two
hours. The Concert Music, composed for the fiftieth anniversary of
the Boston Symphony Orchestra, is more than interesting. It cannot
be called “noble,” not even “grand,” but it holds the attention by its
strength in structure, its spirit, festal without blatancy. For once
there is no too evident desire to stun the hearer. It is as if the
composer had written for his own pleasure. It is virile music with
relieving passages—few in number—that have genuine and simple
beauty of thought and expression; exciting at times by the 162
rushing rhythm.
Hindemith, at the age of eleven, played the viola in the theater and
in the moving-picture house; when he was thirteen, he was a viola
virtuoso, and he now plays in public his own concertos for that
instrument. When he was twenty, he was first concert master of the
Frankfort opera house. His teachers in composition were Arnold
Mendelssohn and Bernhard Sekles at the Hoch Conservatory in
Frankfort. He is the viola player in the Amar Quartet (Licco Amar,
Walter Casper, Paul Hindemith, and Maurits Frank—in 1926 his
brother Rudolf was the violoncellist).

Apropos of a performance of one of his works, in Berlin, the late

Adolf Weissmann wrote in a letter to the Christian Science Monitor:
“Promising indeed among the young German composers is Paul
Hindemith. More than promising he is not yet. For the viola player
Paul Hindemith, travelling with the Amar Quartet through half
Europe, has seldom time enough to work carefully. The greater part
of his compositions were created in the railway car. Is it, therefore,
to be wondered at that their principal virtue lies in their rhythm? The
rhythm of the rolling car is, apparently, blended with the rhythm
springing from within. It is always threatening to outrun all the other
values of what he writes. For that these values exist cannot be
denied.”

A foreign correspondent of the London Daily Telegraph, having heard

one of his compositions, wrote: “It was all rather an exhilarating
nightmare, as if Hindemith had been attempting to prove the
theorem of Pythagoras in terms of parallelograms, which is amusing,
but utterly absurd.”

It has been said by A. Machabey that Hindemith has been influenced

in turn by Wagner, Brahms—“an influence still felt”; Richard Strauss;
Max Reger, who attracted him by his ingenuity and freedom from
elementary technic; Stravinsky, who made himself felt after the war;
and finally by the theatrical surroundings in which he lives. “He is
opposed to post-romanticism. Not being able to escape from
romanticism in his youth, today he seems to be completely stripped
of it. Freed from the despotism of a text, from the preëstablished
plan of programme music, from obedience to the caprices 163
and emphasis of sentiment, music in itself suffices.... The
reaction against romanticism is doubled by a democratic spirit which
was general in Germany after the war.” Therefore he has had many
supporters, who welcomed, “besides this new spirit, an unexpected
technic, unusual polyphony and instrumentation, in which one found
a profound synthesis of primordial rhythms, tonalities enriched and
extended by Schönberg and Hauer, economical and rational
groupings of jazz.” Then his compositions are so varied: chamber
music for the ultra-fastidious; melodies for amateurs; dramatic works
for opera-goers; orchestral pieces for frequenters of concerts; he has
written for débutantes and children; for the cinema, marionettes,
mechanical pianos, brass bands. Work has followed work with an
amazing rapidity.

164
 ARTHUR
HONNEGER
(Born at Havre, France, on March 10, 1892)
 “PACIFIC 231,” ORCHESTRAL MOVEMENT

Some say that Honegger had no business to summon a locomotive

engine for inspiration. No doubt this music of Honegger’s is “clever,”
but cleverness in music quickly palls. Louis Antoine Jullien years ago
in this country excited wild enthusiasm by his Firemen’s Quadrille, in
which a conflagration, the bells, the rush of the firemen, the
squirting and the shout of the foreman, “Wash her, Thirteen!” were
graphically portrayed.

But there is majestic poetry in great machines, even in railway

engines. One of Turner’s most striking pictures is the one depicting a
hare running madly across a viaduct with a pursuing locomotive in
rain and mist. What was the most poetic thing of the Philadelphia
exposition of 1876? The superb Corliss engine, epic in strength and
grandeur. Walt Whitman, Kipling, and others have found inspiration
in a locomotive; why reproach a composer for attempting to express
“the visual impression and the physical sensation” of it? One may
like or dislike Pacific 231, but it is something more than a musical
joke; it was not merely devised for sensational effect.

When Pacific 231 was first performed in Paris at Koussevitzky’s

concerts, May 8 and 15, 1924, Honegger made this commentary:

“I have always had a passionate love for locomotives. To me they—

and I love them passionately as others are passionate in their 165
love for horses or women—are like living creatures.
“What I wanted to express in the Pacific is not the noise of an
engine, but the visual impression and the physical sensation of it.
These I strove to express by means of a musical composition. Its
point of departure is an objective contemplation: quiet respiration of
an engine in state of immobility; effort for moving; progressive
increase of speed, in order to pass from the ‘lyric’ to the pathetic
state of an engine of three hundred tons driven in the night at a
speed of one hundred and twenty per hour.

“As a subject I have taken an engine of the ‘Pacific’ type, known as

‘231,’ an engine for heavy trains of high speed.”

Other locomotive engines are classified as “Atlantic,” “Mogul.” The

number 231 here refers to the number of the “Pacific’s” wheels 2—3
—1.

“On a sort of rhythmic pedal sustained by the violins is built the

impressive image of an intelligent monster, a joyous giant.”

Pacific 231 is scored for piccolo, two flutes, two oboes, English horn,
two clarinets, bass clarinet, two bassoons, double bassoon, four
horns, three trumpets, three trombones, bass tuba, snare drum,
bass drum, cymbals, tam-tam, strings.

The locomotive engine has been the theme of strange tales by

Dickens, Marcel Schwob, Kipling, and of Zola’s novel, La Bête
humaine. It is the hero of Abel Gance’s film, Roué for which it is said
Honegger adapted music, and the American film, The Iron Horse.

166