You are on page 1of 53

Mathematical Foundations of Time

Series Analysis A Concise Introduction


1st Edition Jan Beran
Visit to download the full and correct content document:
https://textbookfull.com/product/mathematical-foundations-of-time-series-analysis-a-c
oncise-introduction-1st-edition-jan-beran/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

The Analysis of Time Series: An Introduction with R 7th


Edition Chris Chatfield

https://textbookfull.com/product/the-analysis-of-time-series-an-
introduction-with-r-7th-edition-chris-chatfield/

Time series analysis 1st Edition Palma

https://textbookfull.com/product/time-series-analysis-1st-
edition-palma/

Design and analysis of time series experiments 1st


Edition Bartos

https://textbookfull.com/product/design-and-analysis-of-time-
series-experiments-1st-edition-bartos/

Nonlinear Time Series Analysis with R 1st Edition Ray


Huffaker

https://textbookfull.com/product/nonlinear-time-series-analysis-
with-r-1st-edition-ray-huffaker/
Applied time series analysis with R Second Edition
Elliott

https://textbookfull.com/product/applied-time-series-analysis-
with-r-second-edition-elliott/

A Concise Geologic Time Scale 1st Edition J G Ogg

https://textbookfull.com/product/a-concise-geologic-time-
scale-1st-edition-j-g-ogg/

ECG Time Series Variability Analysis: Engineering and


Medicine 1st Edition Cornforth

https://textbookfull.com/product/ecg-time-series-variability-
analysis-engineering-and-medicine-1st-edition-cornforth/

Mathematical Foundations of Computer Science 1st


Edition Satyanarayana

https://textbookfull.com/product/mathematical-foundations-of-
computer-science-1st-edition-satyanarayana/

Indexation and Causation of Financial Markets


Nonstationary time series analysis method 1st Edition
Yoko Tanokura

https://textbookfull.com/product/indexation-and-causation-of-
financial-markets-nonstationary-time-series-analysis-method-1st-
edition-yoko-tanokura/
Jan Beran

Mathematical
Foundations
of Time Series
Analysis
A Concise Introduction
Mathematical Foundations of Time Series Analysis
Jan Beran

Mathematical Foundations
of Time Series Analysis
A Concise Introduction

123
Jan Beran
Department of Mathematics and Statistics
University of Konstanz
Konstanz, Germany

ISBN 978-3-319-74378-3 ISBN 978-3-319-74380-6 (eBook)


https://doi.org/10.1007/978-3-319-74380-6

Library of Congress Control Number: 2018930982

Mathematics Subject Classification (2010): 62Mxx, 62M10

© Springer International Publishing AG, part of Springer Nature 2017


This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by the registered company Springer International Publishing AG part
of Springer Nature.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The historical development of time series analysis can be traced back to many
applied sciences, including economics, meteorology, physics or communications
engineering. Theoretical developments of the subject are closely linked to progress
in the mathematical theory of stochastic processes and mathematical statistics.
There are a number of excellent books on time series analysis, including Grenander
and Rosenblatt (1957), Box and Jenkins (1970), Hannan (1970), Anderson (1971),
Koopmans (1974), Fuller (1976), Priestley (1981), Brockwell and Davis (1991),
Hamilton (1994), Diggle (1996), Brillinger (2001), Chatfield (2003), Lütkepohl
(2006), Durbin and Koopmann (2012), Woodward et al. (2016), and Shumway and
Stoffer (2017).
Time series analysis is now a well-established scientific discipline with rigorous
mathematical foundations. On the other hand, it is a very broad subject area, and,
due to the diverse sciences that contributed to its development, the time series
vocabulary is permeated with terminology reflecting the diversity of applications
(cf. Priestley 1981, Preface p. vii). This book is an attempt to summarize some of the
main principles of time series analysis, with the hope that the concise presentation is
helpful for teaching students with a mathematical background. The book grew out
of lectures taught to students of mathematics, mathematical finance, physics and
economics at the University of Konstanz.
I would like to thank Martin Schützner, Mark Heiler, Dieter Schell, Evgeni
Shumm, Nadja Schumm, Arno Weiershäuser, Dirk Ocker, Karim Djaidja, Haiyan
Liu, Britta Steffens, Klaus Telkmann, Yuanhua Feng, Philipp Sibbertsen, Bikramjit
Das, Rafal Kulik, Liudas Giraitis, Sucharita Ghosh and other colleagues for fruitful
collaboration; and to Volker Bürkel for reading parts of a preliminary manuscript.
Thanks go also to the University of Konstanz for granting me a sabbatical with
the purpose of working on this book. Most importantly, I would like to thank my
family, Céline, Sucharita and Sir Hastings—our Coton de Tuléar—for keeping me
motivated.

Konstanz, Germany Jan Beran


November 2017

v
Contents

1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.1 What Is a Time Series?. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 Time Series Versus iid Data. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 2
2 Typical Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5
2.1 Fundamental Properties . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5
2.1.1 Ergodic Property with a Constant Limit . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Strict Stationarity . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
2.1.3 Weak Stationarity . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8
2.1.4 Weak Stationarity and Hilbert Spaces. . .. . . . . . . . . . . . . . . . . . . . 11
2.1.5 Ergodic Processes . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 32
2.1.6 Sufficient Conditions for the a.s. Ergodic Property
with a Constant Limit . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 34
2.1.7 Sufficient Conditions for the L2 -Ergodic Property
with a Constant Limit . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
2.2 Specific Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
2.2.1 Gaussian Processes . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
2.2.2 Linear Processes in L2 .˝/. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 40
2.2.3 Linear Processes with E.Xt2 / D 1 . . . . . .. . . . . . . . . . . . . . . . . . . . 44
2.2.4 Multivariate Linear Processes . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 48
2.2.5 Invertibility .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 49
2.2.6 Restrictions on the Dependence Structure . . . . . . . . . . . . . . . . . . 63
3 Defining Probability Measures for Time Series . . . . .. . . . . . . . . . . . . . . . . . . . 69
3.1 Finite Dimensional Distributions .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 69
3.2 Transformations and Equations .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 70
3.3 Conditions on the Expected Value .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
3.4 Conditions on the Autocovariance Function . . . .. . . . . . . . . . . . . . . . . . . . 73
3.4.1 Positive Semidefinite Functions .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 73
3.4.2 Spectral Distribution .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 77
3.4.3 Calculation and Properties of F and f . . .. . . . . . . . . . . . . . . . . . . . 86

vii
viii Contents

4 Spectral Representation of Univariate Time Series . . . . . . . . . . . . . . . . . . . . . 101


4.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 101
4.2 Harmonic Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 102
4.3 Extension to General Processes. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 105
4.3.1 Stochastic Integrals with Respect to Z . .. . . . . . . . . . . . . . . . . . . . 105
4.3.2 Existence and Definition of Z . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 112
4.3.3 Interpretation of the Spectral Representation .. . . . . . . . . . . . . . 122
4.4 Further Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 122
4.4.1 Relationship Between Re Z and Im Z . .. . . . . . . . . . . . . . . . . . . . 122
4.4.2 Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 123
4.4.3 Overtones .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 124
4.4.4 Why Are Frequencies Restricted to the Range Œ; ? . . . 125
4.5 Linear Filters and the Spectral Representation... . . . . . . . . . . . . . . . . . . . 129
4.5.1 Effect on the Spectral Representation .. .. . . . . . . . . . . . . . . . . . . . 129
4.5.2 Elimination of Frequency Bands . . . . . . . .. . . . . . . . . . . . . . . . . . . . 134
5 Spectral Representation of Real Valued Vector Time Series. . . . . . . . . . . 137
5.1 Cross-Spectrum and Spectral Representation .. .. . . . . . . . . . . . . . . . . . . . 137
5.2 Coherence and Phase .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 146
6 Univariate ARMA Processes . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 161
6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 161
6.2 Stationary Solution .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 161
6.3 Causal Stationary Solution .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 166
6.4 Causal Invertible Stationary Solution . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 169
6.5 Autocovariances of ARMA Processes . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 170
6.5.1 Calculation by Integration . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 170
6.5.2 Calculation Using the Autocovariance Generating
Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 170
6.5.3 Calculation Using the Wold Representation .. . . . . . . . . . . . . . . 175
6.5.4 Recursive Calculation .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 176
6.5.5 Asymptotic Decay . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 177
6.6 Integrated, Seasonal and Fractional ARMA and ARIMA
Processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 185
6.6.1 Integrated Processes . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 185
6.6.2 Seasonal ARMA Processes . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 186
6.6.3 Fractional ARIMA Processes . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 187
6.7 Unit Roots, Spurious Correlation, Cointegration . . . . . . . . . . . . . . . . . . . 200
7 Generalized Autoregressive Processes .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 203
7.1 Definition of Generalized Autoregressive Processes . . . . . . . . . . . . . . . 203
7.2 Stationary Solution of Generalized Autoregressive Equations .. . . . 204
7.3 Definition of VARMA Processes . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209
7.4 Stationary Solution of VARMA Equations . . . . .. . . . . . . . . . . . . . . . . . . . 211
7.5 Definition of GARCH Processes . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 213
7.6 Stationary Solution of GARCH Equations .. . . . .. . . . . . . . . . . . . . . . . . . . 214
Contents ix

7.7 Definition of ARCH(1) Processes. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 219


7.8 Stationary Solution of ARCH(1) Equations . . .. . . . . . . . . . . . . . . . . . . . 220
8 Prediction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
8.1 Best Linear Prediction Given an Infinite Past . . .. . . . . . . . . . . . . . . . . . . . 223
8.2 Predictability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 225
8.3 Construction of the Wold Decomposition from f . . . . . . . . . . . . . . . . . . . 230
8.4 Best Linear Prediction Given a Finite Past . . . . . .. . . . . . . . . . . . . . . . . . . . 235
9 Inference for ,  and F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 241
9.1 Location Estimation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 241
9.2 Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 244
9.3 Nonparametric Estimation of  . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 253
9.4 Nonparametric Estimation of f . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 262
10 Parametric Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 281
10.1 Gaussian and Quasi Maximum Likelihood Estimation .. . . . . . . . . . . . 281
10.2 Whittle Approximation . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 284
10.3 Autoregressive Approximation . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 287
10.4 Model Choice.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 289

References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 293

Author Index.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 299

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 303


Chapter 1
Introduction

1.1 What Is a Time Series?

Definition 1.1 Let k 2 N, T  R. A function

x W T ! Rk , t ! x t

or, equivalently, a set of indexed elements of Rk ,


˚ 
xt jxt 2 Rk ; t 2 T

is called an observed time series. We also write

xt (t 2 T) or .xt /t2T :

Definition 1.2 Let k 2 N, T  R,


 T
˝ D Rk D space of functions X W T ! Rk ;

F D -algebra on ˝;

P D probability measure on .˝; F / :

The probability space .˝; F ; P/, or equivalently the set of indexed random
variables
˚ 
Xt jXt 2 Rk ; t 2 T , .Xt /t2T  P

© Springer International Publishing AG, part of Springer Nature 2017 1


J. Beran, Mathematical Foundations of Time Series Analysis,
https://doi.org/10.1007/978-3-319-74380-6_1
2 1 Introduction

Table 1.1 Types of time series Xt 2 Rk (t 2 T)


Property Terminology
kD1 Univariate time series
k2 Multivariate time series
T countable, 8a < b 2 R W T \ Œa; b finite Discrete time
T discrete, 9u 2 RC s.t. tjC1  tj D u Equidistant time
T D Œa; b (a < b 2 R), T D RC or T D R Continuous time

is called a time series, or time series model. Instead of .˝; F ; P/ we also write

Xt (t 2 T) or .Xt /t2T :

Moreover, for a specific realization ! 2 ˝, we write Xt .!/ and

.xt /t2T D .Xt .!//t2T D sample path of .Xt /t2T ;

.xti /iD1;:::;n D .Xti .!//iD1;:::;n D finite sample path of Xt :

Remark 1.1 ˝ may be more general than in Definition 1.2. Similarly, the index set
T may be more general than a subset of R, but it must be ordered and metric. Thus,
.Xt /t2T is a stochastic process with an ordered metric index set T:
Remark 1.2 An overview of the most common types of time series Xt 2 Rk (t 2 T,
T ¤ ;) is given in Table 1.1.
Remark 1.3 If Xt Dequidistant time series, then we may set w.l.o.g. T  Z.

1.2 Time Series Versus iid Data

What distinguishes statistical analysis of iid data from time series analysis? We
illustrate the question by considering the case of equidistant univariate real valued
time series Xt 2 R (t 2 Z).
Problem 1.1 Is consistent estimation of P possible?
Solution 1.1 The answer depends on available a priori information and assump-
tions one is willing to make. This is illustrated in the following.
1.2 Time Series Versus iid Data 3

Notation 1

FXt .x/ D P .Xt  x/ D marginal distribution function of Xt at time t

1X
n
Fn .x/ D 1 fxt  xg D empirical marginal distribution function
n tD1

Lemma 1.1 Assumption:

Xt 2 R (t 2 Z) iid.

Then

8t 2 Z: FXt D FX0

and

P is fully specified by FX0 :

Theorem 1.1 (Glivenko-Cantelli-Theorem) Assumption:

Xt 2 R (t 2 Z) iid.

Then
 
P lim sup j Fn .x/  FX0 .x/j D 0 D 1:
n!1 x2R

Proof See e.g. van der Vaart (1998). For the original proofs see Glivenko (1933)
and Cantelli (1933).
Conclusion 1 Given the iid assumption, P can be estimated consistently. No
additional assumptions are required.
This is not true in general under non-iid assumptions.
Example 1.1

Xt D t C Zt (t 2 Z), Zt iid N .0; 1/

t (t 2 Z) unknown.
4 1 Introduction

Consistent estimation of t is not possible, unless additional assumptions on t are


imposed.
Example 1.2

Xt D U (t 2 Z),

0 < p D P .U D 1/ D 1  P .U D 0/ < 1:

Consistent estimation of p is not possible.


Conclusion 2 In general, consistent estimation of P is not possible, unless addi-
tional assumptions are imposed. The problem is:

observed time series D .x1 ; : : : ; xn /


D .X1 .!/ ; : : : ; Xn .!//
D sample of size one from n-dimensional distribution.
¤ sample from the infinite dimensional distribution P on RZ
Chapter 2
Typical Assumptions

In general, consistent estimation of P is not possible without additional assumptions.


In this chapter, typical assumptions used in time series analysis are discussed. For
simplicity we focus on equidistant univariate real valued time series Xt 2 R (t 2 Z).

2.1 Fundamental Properties

2.1.1 Ergodic Property with a Constant Limit

The asymptotic distribution of many statistics follows—sometimes after a compli-


cated proof or suitable transformations—from the asymptotic distribution of sums.
Notation 2

X
n
xN D xN n D n1 Xt
tD1

Definition 2.1 Xt 2 R (t 2 Z) has the almost sure ergodic property with a constant
limit, or a.s. EPCL, if
 
9 2 R s.t. P lim xN D  D 1:
n!1

Sometimes one also calls this the mean-ergodic property in the a.s. sense.

© Springer International Publishing AG, part of Springer Nature 2017 5


J. Beran, Mathematical Foundations of Time Series Analysis,
https://doi.org/10.1007/978-3-319-74380-6_2
6 2 Typical Assumptions

Definition 2.2 Xt 2 R (t 2 Z) has the L2 -ergodic property with a constant limit, or


L2 -EPCL, if
h i
9 2 R s.t. lim E .Nx  /2 D 0:
n!1

Sometimes one also calls this the mean-ergodic property in the L2 -sense.
Under which circumstances can it happen that the EPCL does not hold? Three
main problems can occur, as outlined in the following.
Problem E1 Lack of stability: Distribution of XtC1 ; XtC2 ; : : : ; XtCn changes too
much as a function of t so that .xt /tD1;:::;n is not sufficiently representative for P
on RZ .
Example 2.1

"t iid, E."t / D 0, "2 D var."t / < 1;

Xt D ˇt C "t :

Then

1 n .n C 1/
xN D ˇ C "N;
n 2

P .jNxj ! 1/ D 1:

Example 2.2

"t iid, E."t / D 0, "2 < 1;

X
t
Xs D 0 (s  0), Xt D "s (t  1).
sD1

Then

X
n
xN D n1 t"ntC1 ;
tD1

"2 X 2
n
var.Nx/ D t ! 1:
n2 tD1

Problem E2 High variability of the marginal distribution FXt .


2.1 Fundamental Properties 7

Example 2.3 For Cauchy distributed iid Xt , we have

xN D X1 :
d

Problem E3 Absorbing states:

Xt (t 2 Z) Ft -measurable,

9t 2 Z, At 2 Ft s.t. 0 < P .Xt 2 At / < 1 and P .8s > t W Xs 2 At jAt / D 1:

Then

At D absorbing state.

Example 2.4

Xt D U (t 2 Z),

0 < p D P .U D 1/ D 1  P .U D 0/ < 1:

Then

At D fXt D 1g D absorbing state.

2.1.2 Strict Stationarity

Definition 2.3 Xt 2 R (t 2 Z) is called strictly stationary or strongly stationary, if

8k 2 Z; 8m 2 N; 8t1 ; : : : ; tm 2 Z: .Xt1 ; : : : ; Xtm / D .Xt1 Ck ; : : : ; Xtm Ck / :


d

Example 2.5 Xt 2 R (t 2 Z) iid is strictly stationary.


Example 2.6

X
q
Xt D j "t1 (t 2 Z),
jD0

"t 2 R (t 2 Z) iid, j 2 R (j D 0; : : : ; q)

is strictly stationary. Xt is called a moving average process of order q, or MA(q)


process.
8 2 Typical Assumptions

Remark 2.1 Strict stationarity solves Problem E1, but not Problems E2 and E3.

2.1.3 Weak Stationarity

Definition 2.4 Let

Xt 2 R (t 2 Z) s.t. 8t 2 Z: E .jXt j/ < 1:

Then

t D E .Xt / (t 2 Z)

is called the expected value function, or mean function, of Xt . If


 
E Xt2 < 1;

then

 WZZ!R

with

 .s; t/ D cov .Xs ; Xt / D E Œ.Xs  s / .Xt  t /

is called the autocovariance function (acf) of Xt , and

 .s; t/
 .s; t/ D corr .Xt ; Xs / D p
 .s; s/  .t; t/

is called the autocorrelation function (acf) of Xt .


Remark 2.2

 .t; t/ D var .X t / ,  .t; t/ D 1

Remark 2.3 For

Xt 2 C (t 2 Z),

we define
h i
 .s; t/ D cov .Xs ; Xt / D E .Xs  s / .Xt  t / :
2.1 Fundamental Properties 9

Lemma 2.1

 .t; s/ D  .s; t/

Proof
h i h i
 .s; t/ D E .Xs  s / .Xt  t / D E .Xs  s / .Xt  t /
h i
D E .Xt  t / .Xs  s / D  .t; s/

Definition 2.5 Xt 2 R (t 2 Z) is called second order stationary, or weakly


stationary, if
 
E Xt2 < 1;

9 2 R s.t. 8t 2 Z W E .Xt / D ;

9 W Z ! R s.t. 8s; t 2 Z W cov .Xs ; Xt / D  .t  s/ :

Lemma 2.2

weak stationarity ¼ strong stationarity

Proof Counterexample for »:


2
Z2iC1 1
X2i D Z2i (i 2 Z), X2iC1 D p (i 2 Z)
2

where

Zi (i 2 Z) iid N .0; 1/ -variables

Counterexample for º:

Xt (t 2 Z) iid, Cauchy distributed

Lemma 2.3 Assumptions:

Xt 2 R (t 2 Z) strictly stationary, E.Xt2 / < 1


10 2 Typical Assumptions

Then

Xt (t 2 Z) weakly stationary

Proof
a) : The Cauchy-Schwarz inequality implies

E2 .jXt j/  E Xt2 < 1;

and hence

9t D E .Xt / 2 R.

Thus, together with strong stationarity

8s; t 2 Z W s D t D  2 R

b)  : The Cauchy-Schwarz inequality implies


   
E2 .Xs Xt /  E2 .jXs j jXt j/  E Xs2 E Xt2 < 1:

Together with strong stationarity we then have

8t; k 2 Z W E .Xt XtCk / D E .X0 Xk / ;

and

cov .Xt ; XtCk / D cov .X0 ; Xk / D E .X0 Xk /  2 D  .k/ :

  Problem E1 w.r.t. first two moments of Xt , and


Remark 2.4 Weak stationarity solves
Problem E2 in the sense that E Xt2 < 1. It does not solve Problem E3.
Example 2.7

Xt D U (t 2 Z)

where

0 < p D P .U D 1/ D 1  P .U D 0/ < 1;

is weakly and strictly stationary, but

At D fXt D 1g D absorbing state.


2.1 Fundamental Properties 11

2.1.4 Weak Stationarity and Hilbert Spaces

For an introduction to Hilbert spaces see e.g. Young (1988).


Definition 2.6

R D space of R-valued random variables on probability space .˝; F ; P/ ;


˚   
L2 .˝/ D L2 .˝; R/ D XjX 2 R s.t. E X 2 < 1 ;

hX; Yi D hX; YiL2 .˝/ D E .XY/ (X; Y 2 L2 .˝/ ),

kXk2L2 .˝/ D hX; Xi (X 2 L2 .˝/ ).

Remark 2.5

kXk2L2 .˝/ D var .X/ C E2 .X/

Remark 2.6 Analogous definitions for

C D space of C-valued random variables on .˝; F ; P/ ;

with
 
hX; Yi D hX; YiL2 .˝:C/ D E XY :

Lemma 2.4 hX; YiL2 .˝:R/ and hX; YiL2 .˝:C/ are scalar products, if we interpret X D
Y as P.X D Y/ D 1.
Proof
1)
       
hX; Yi D E XY D E YX D E YX D E YX D h Y; Xi

2)
   
haX; Yi D E aXY D aE XY D a hX; Yi

3)
   
hX C Y; Zi D E .X C Y/ Z D E XZ C E XZ D hX; Zi C h Y; Zi
12 2 Typical Assumptions

4)
 
hX; Xi D E jXj2  0
 
hX; Xi D E jXj2 D 0 , X D 0 a.s.

Lemma 2.5 L2 .˝; R/ and L2 .˝; C/ are complete.


Proof We need to show the implication:

Xn (n 2 N) Cauchy sequence in L2 .˝/

9random variable X s.t. a) lim kXn  Xk2 D 0 and b) X 2 L2 .˝/ :


n!1

a)

Xn 2 L2 .˝/ (n 2 N) Cauchy sequence , kXn  Xm k ! 0 (m; n ! 1)

Hence,

9subsequence Xnj (j 2 N) s.t. XnjC1  Xnj  2j (j 2 N).

Setting n0 D 0, X0 D 0, we have

X
j
Xnj D .Xni  Xni1 / :
iD1

Define

X
j
Wj D jXni  Xni1 j :
iD1

Then
ˇ j ˇ
ˇ ˇ ˇˇX ˇ
ˇ
ˇXnj ˇ D ˇ .Xni  Xni1 /ˇ  Wj ;
ˇ ˇ
iD1
2.1 Fundamental Properties 13

and therefore

ˇ ˇ   X
j
E ˇXnj ˇ  E Wj D E .jXni  Xni1 j/ :
iD1

The Cauchy-Schwarz inequality implies

E .jXni  Xni1 j/  kXni  Xni1 k ;

so that

ˇ ˇ X j
E ˇXnj ˇ  Wj  kXni  Xni1 k
iD1

X
j
 kX1 k C 2j  kX1 k C 1 < 1:
iD1

Since
 
Wj  0 nondecreasing, E Wj  kX1 kL2 .˝/ C 1 < 1;

we obtain
 
P 9W D lim Wj < 1 D 1
j!1

and hence
 
P 9X D lim Xnj 2 R D 1:
j!1

b)
ˇ ˇ2 !  
ˇ ˇ ˇ ˇ2
kXn  Xk D E ˇˇXn  lim Xnj ˇˇ D E lim ˇXn  Xnj ˇ
2
j!1 j!1
  ˇ
ˇ ˇ2 ˇ2 
ˇ
D E liminf Xn  Xnj ˇ  liminfE ˇXn  Xnj ˇ
j!1 j!1

2
D liminf Xn  Xnj :
j!1
14 2 Typical Assumptions

Now
2
Xn Cauchy sequence ) lim liminf Xn  Xnj D0
n!1 j!1

2
) lim kXn  Xk2  lim liminf Xn  Xnj D 0:
n!1 n!1 j!1

c)

kXk  kX  Xn k C kXn k < 1 ) X 2 L2 .˝/

Corollary 2.1 L2 .˝; R/ and L2 .˝; C/ are Hilbert spaces.


Proof Follows from Lemma 2.5.
Corollary 2.2 Assumption:

Xt 2 R (t 2 Z) weakly stationary.

Then

8t 2 Z W Xt 2 L2 .˝/

Proof Follows from the definition of weak stationarity with E.Xt2 / < 1.
Corollaries 2.1 and 2.2 are useful for linear prediction:
Definition 2.7 Let

Ft D  .Xs ; s  t/ ;
˚ 
Xt D Y j Y 2 L2 .˝/ , Ft -measurable :

Then, for k  1 and Y 2 Xt ,


h i
k2 . Y/ D E .XtCk  Y/2

D k-step prediction mean squared error of Y

Problem 2.1 Find prediction of XtCk given Ft that minimizes k2 .
Solution 2.1 Projection theorem for Hilbert spaces.
2.1 Fundamental Properties 15

Theorem 2.1 (Continuity) Assumptions:

.H; h; i/ Hilbert space,

x; y; xn ; yn 2 H (n 2 N) s.t. lim kxn  xk D 0, lim kyn  yk D 0


n!1 n!1

Then

lim kxn k D kxk


n!1

and

lim hxn ; yn i D hx; yi :


n!1

Proof
a)

kxn k  kxn  xk C kxk

implies

kxn k  kxk  kxn  xk :

Similarly,

kxk  kxn  xk C kxn k

implies

 kxn  xk  kxn k  kxk :

Hence,

0  lim jkxn k  kxkj  lim kxn  xk D 0


n!1 n!1

b)

lim jhxn ; yn i  hx; yij  lim .jhxn ; yn  yij C jhxn  x; yij/


n!1 n!1

 lim kxn k kyn  yk C lim kxn  xk kyk D 0


n!1 n!1
16 2 Typical Assumptions

Theorem 2.2 (Projection Theorem) Assumptions:

.H; h; i/ Hilbert space, x 2 H, A  H, A closed, convex

Then

9ŠOx 2 A s.t. kx  xO k D inf kx  yk :


y2A

Proof
1) Existence: Let

dx D inf kx  yk ;
y2A

1
yn 2 A (n 2 N) s.t. kx  yn k2 < dx2 C :
n
Convexity of A implies

yn C ym
2A
2
and therefore

yn C ym
 x  dx :
2

Now

kyn  ym k2 D k. yn  x/ C .x  ym /k2
D kyn  xk2 C kym  xk2  2 h yn  x; ym  xi

and
2
yn C ym
4 x D k. yn  x/ C . ym  x/k2
2
D kyn  xk2 C kym  xk2 C 2 h yn  x; ym  xi ;

so that
2
yn C ym
4 x C kyn  ym k2 D 2 kyn  xk2 C 2 kym  xk2 :
2
2.1 Fundamental Properties 17

Hence,
2
yn C ym
kyn  ym k2 D 2 kyn  xk2 C 2 kym  xk2  4 x
2
 2 kyn  xk2 C 2 kym  xk2  4dx2
   
1 1 2 2
 2 dx2 C C 2 dx2 C  4dx2 D C ;
n m n m
so that
yn D Cauchy sequence,

9Ox D lim yn 2 AN D A
n!1

and
1
kx  yn k2 < dx2 C :
n
Thus,
xO 2 A and kOx  yk D inf kx  yk :
y2A

2) Uniqueness: Let

xO 1 ; xO 2 s.t. kOx1  yk D kOx2  yk D inf kx  yk D dx :


y2A

Convexity of A implies

xO 1 C xO 2
2A
2
and hence
2
xO 1 C xO 2
x  dx2 :
2

Then
2
xO 1 C xO 2
kOx1  xO 2 k2 D 2 kOx1  xk2 C 2 kOx2  xk2  4 x
2
2
xO 1 C xO 2
D 4dx2  4 x
2
 4dx2  4dx2 D 0:
18 2 Typical Assumptions

so that

kOx1  xO 2 k2 D 0:

Definition 2.8 Let .H; h; i/ be a Hilbert space over R. Then A  H is called a
linear subspace of H, if
0 2 A;
8x; y 2 A W x C y 2 A;

8x 2 A; 2 R W x 2 A:

An analogous definition applies for H over C.


Definition 2.9

A? D fx j x 2 H s.t. 8y 2 A W hx; yi D 0g
D orthogonal complement of A
We write

A ? B , 8x 2 A; 8y 2 B W hx; yi D 0:

Lemma 2.6

.H; h; i/ Hilbert space over R or C, A  H linear subspace

) A convex

Proof Let
X
c1 ; : : : ; ck s.t. ci D 1, x1 ; : : : ; xk 2 A:

Then

X
k
A linear subspace ) ci xi 2 A:
iD1

Hence A is convex.
Lemma 2.7

.H; h; i/ Hilbert space, A  H

) A? closed linear subspace


2.1 Fundamental Properties 19

Proof
a)

8x 2 A W h0; xi D 0 ) 0 2 A?

b)

y; z 2 A? ) 8x 2 A W h y C z; xi D h y; xi C hz; xi D 0 ) y C z 2 A?

c)

y 2 A? ) 8x 2 A W h y; xi D h y; xi D 0 ) y 2 A?

Thus, a), b), c) imply

A D linear subspace.

d) Let

yn 2 A? ; y 2 H s.t. lim kyn  yk D 0:


n!1

Then

8x 2 A W h y; xi D lim h yn ; xi D 0 ) y 2 A? ) A? closed
n!1

Corollary 2.3 Assumptions:

.H; h; i/ Hilbert space, x 2 H, A D closed linear subspace, xO 2 A

Then

kx  xO k D inf kx  yk , x  xO 2 A?
y2A

Proof
1) “(” Let

xO 2 A s.t. x  xO 2 A?

and

y 2 A arbitrary.
20 2 Typical Assumptions

Then

A linear space ) xO  y 2 A

and

x  xO 2 A? ) kx  yk2 D kx  xO k2 C kOx  yk2  kx  xO k2

so that

kx  xO k2 D inf kx  yk2 :
y2A

2) “)” Follows from “(” and uniqueness.


Definition 2.10 Let A and xO be as in Corollary 2.3. Then

PA W H ! A, x ! PA .x/ D xO

is called (orthogonal) projection mapping, or projection operator on A.


Definition 2.11 Let Xt 2 R (t 2 Z) be weakly stationary,

Ft D  .Xs ; s  t/ ;
˚ 
Xt D Y j Y 2 L2 .˝/ , Ft -measurable ;

k 2 N, XO tCk 2 Xt :

Then

XO tCk D optimal forecast of XtCk given Ft (or given Xs , s  t)

,
 
k2 XO tCk D inf k2 . Y/ :
Y2Xt

Definition 2.12 Let

.˝; F ; P/ D probability space,

X W ˝ ! R s.t. X is F -measurable,

G  F s.t. G D -algebra.
2.1 Fundamental Properties 21

A G -measurable random variable Y is called conditional expected value of X given


G , if
Z Z
8A 2 G W YdP D XdP:
A A

We then write

Y D E .X j G / :

Remark 2.7 If E.X/ exists, then E.XjG / exists and is unique in the sense of almost
sure equality (see e.g. Kallenberg 2002).
Corollary 2.4 Assumptions:

Xt 2 R (t 2 Z) weakly stationary,

XO tCk 2 Xt :

Then
 
k2 XO tCk D inf k2 . Y/ , XO tCk D E .XtCk j Ft / :
Y2Xt a:s:

Proof

Xt ; XtCk 2 L2 .˝/ , Xt D closed convex subset

Then, for XO tCk 2 Xt ,

XtCk  XO tCk D inf kXtCk  Yk , XO tCk 2 Xt and XtCk  XO tCk 2 Xt ?


Y2Xt

and this random variable is unique. Now

8Z 2 Xt W E ŒZE .XtCk j Ft / D E ŒZXtCk 

so that

8Z 2 Xt W hZ; XtCk  E .XtCk j Ft /i D 0;

and therefore

kXtCk  E .XtCk j Ft /k D inf kXtCk  Yk :


Y2Xt
Another random document with
no related content on Scribd:
manier waarop de ouders met hun jongen omgingen, maakte, dat
Piet nooit achterbaks behoefde te zijn.

Toen mijnheer uit zijn studeerkamer kwam, kon zijn vrouw niet
nalaten hem te vragen, of hij niets bizonders aan Piet had
opgemerkt.

„Aan Piet!” zei mijnheer heel verbaasd. „De jongen was net als
anders. Wat haal je je nu weer voor muizenissen in ’t hoofd?”

Het antwoord bevredigde mevrouw maar matig en zonder er verder


op door te gaan, bleven haar gedachten nog lang bij haar jongen
verwijlen.

En zoo gebeurde het, dat, toen reeds alles in diepe rust was, zij nog
met open oogen lag en den slaap niet kon vatten.

Ze had de klok reeds elf en half twaalf hooren [104]slaan, keerde zich
om en om en voelde zich hoogst onrustig.

Plotseling hoort zij in de stilte van den nacht een geluid als het
piepen van een raam dat opengeschoven wordt.

Even denkt ze nog dat het bij de buren is, doch dan, als gedreven
door een voorgevoel, staat ze haastig op, en begeeft zich ijlings naar
Piet’s kamertje.

Ze opent de deur.…! Daar staan moeder en zoon tegenover elkaar.


Piet, met jas aan en pet op, juist gereed om het raam uit te klimmen,
blijft als versteend staan.

Zijn moeder kijkt hem met een mengeling van verbazing en verdriet
aan en dit stille kijken treft den jongen dieper dan hevige verwijten
hadden kunnen doen.
En Piet, het Opperhoofd der Inka’s, de ontdekker van het Hol van
Kaan, barst in snikken uit, terwijl hij, net als toen hij nog klein was,
z’n hoofd tegen haar schouder aan verbergt.

„Wàt wilde je doen, jongen, nu mag je ’t me toch vertellen,” vraagt ze


zacht.

Ze gaat op den rand van zijn bed zitten en Piet vertelt haar met
horten en stooten het geheele plan.

Hij ziet er zóó ontdaan en verdrietig uit, dat moeder medelijden met
hem heeft, want ze voelt heel goed, wat ’t voor hem is, zijn makkers
te verraden.

„Dàt was ’t dus, Piet. Ja, moeder voelde wel, [105]dat er iets gaande
was. Ik wil je verder geen verwijten maken, alleen wil ik je zeggen,
dat je een domme, roekelooze jongen bent geweest. Want, denk je
er eens even in, dat ik wat later was gekomen en je bed leeg had
gevonden. Kun je voelen, vent, hoe ’n radelooze angst je ons dan
bezorgd had?”

Weer ging Piet’s hoofd schuil in moeder’s schouder.

„En luister nu eens, jongen; diezelfde angst kunnen vannacht de


ouders van drie jongens hebben. Mogen we dit toelaten?”

Bij die woorden kijkt Piet verschrikt op.

„Niet ze verraden,” wil hij zeggen. Maar moeder kijkt hem ernstig in
de oogen en zegt:

„En als er nu vannacht, eens een ongeluk met een van je vriendjes
gebeurt, Piet!… Denk eens aan Bob van Eest …”
„Dat is waar, moeder,” zegt Piet haastig. „Ze mogen niet gaan. Ik ga
’t ze zeggen.”

„Neen baas, dat zullen we vader laten doen. ’t Is wel hard, om hem
uit z’n rust te halen, maar ’t is noodig.”

In minder dan geen tijd is mijnheer Kaan in de kleeren. Het was een
heel werk geweest, hem uit z’n vasten slaap te halen en hem in
korte trekken de toedracht der zaak duidelijk te maken.

Gelukkig nam hij het geval nogal niet te zwaar op. Piet kreeg een
voorloopigen uitbrander met het bevel onmiddellijk naar bed te gaan.

En toen toog mijnheer er in den nacht op uit [106]om de drie andere


roekelooze knapen op het rechte pad te brengen.

Het is precies vijf minuten over twaalven als mijnheer op den


Nieuwen Binnenweg aankomt. Even vóór het huis waar Ambro
woont, ziet hij drie kleine, donkere gedaanten heen en weer loopen.

„De schelmen,” bromt mijnheer in zichzelf. „Ze hebben ’t dan toch


gewaagd.”

Hij besluit hen achterop te loopen. De jongens hebben er geen flauw


idee van, dat ze bespied worden. Ambro, die tusschen Puckie en
Chris loopt, raadpleegt z’n horloge en zegt ongeduldig:

„Piet komt niet, hoor! Dat valt me van hem tegen, hij is toch anders
altijd van de partij.”

„Laten we een steentje tegen z’n raam gooien, misschien is ie in


slaap gevallen,” zegt Chris.

„De wekker zal ie niet hebben laten afloopen,” grinnikt Puckie.


De laatste woorden werden gehoord door den heer Kaan, die het
drietal zachtjes is genaderd.

„Zoo heeren,” klinkt z’n stem tamelijk hard door de stille straat. „Nog
zóó laat op ’t pad?”

De drie nachtbrakers blijven als versteend staan.

„Waar gaat de tocht heen?”

„Wandelen naar Overschie,” zei Ambro, die als altijd de moedigste


was.

„Een vreemd uur om zoo’n wandeling te maken. Weten jullie ouders


daar wel van?”

„Natuurlijk niet, mijnheer, u begrijpt, dat we dan zooiets niet zouden


mogen doen. Die slapen [107]nu allen gerust en hebben geen flauw
idee, dat we niet in ons bed liggen.”

„Behalve de moeder van Piet,” zegt mijnheer. „Die slaapt niet


gerust.”

„Nou mijnheer,” zegt Chris. „Piet is toch niet meegegaan.”

„Neen, dank zij de oplettendheid van zijn moeder, die hem op


heeterdaad betrapte toen hij al met één been buiten het raam stond.”

„Zie je wel,” verdedigde Puckie, „Piet heeft ons niet verraden.”

Hiermede gaf hij eigenlijk een antwoord op de gedachten van alle


drie.

„Neen jongens, je begrijpt ’t verkeerd. Over Piet is m’n vrouw niet


meer in onrust, wel over drie andere roekelooze bengels.”
Het begon tot de jongens door te dringen en ze keken verlegen naar
den grond. Ambro schopte denkbeeldige steentjes van de straat
weg.

Er volgde een pijnlijke stilte, die eindelijk onderbroken werd door


meneer Kaan.

„Nu, wat denken jullie te doen? Wordt het plan doorgezet?”

„We zijn stom geweest, meneer,” bekende Ambro volmondig. „We


hebben aan alles wat u daar zegt, niet gedacht.”

„Wie weet, of mijn moeder nou ook niet in onrust zit,” zei Puckie,
wiens geweten begon te spreken.

„Als ik jullie een goeden raad mag geven, kruip dan weer net zoo
zacht je bed in, als je er uitgekomen [108]bent en vertel morgen aan
je ouders wat je hadt willen doen. Ik hoop, dat jullie in het vervolg
verstandiger zult zijn en ik vertrouw jullie nu verder.”

Meneer reikt ze daarop alle drie de hand en de druk van die drie
jongenshanden bewijst hem, dat hij naar hun hart gesproken heeft
en dergelijke grapjes niet weer zullen voorkomen.

Mijnheer verdwijnt in den donkeren nacht en haast zich huiswaarts,


teneinde niets meer te verliezen van zijn kostbare nachtrust.

De jongens gaan op een drafje naar huis en ’t gelukte hen allen even
vlug hun slaapvertrek te bereiken als ze dit verlaten hadden.

Den volgenden morgen, aan het ontbijt, vertelde Ambro de


geschiedenis aan zijn ouders, die het eerst niet gelooven wilden en
hem zeiden: „je hebt vast gedroomd.”
Maar toen het dienstmeisje, dat de thee binnenbracht, met een
verbaasd gezicht vertelde, dat vanmorgen de knippen van de
straatdeur waren en het raampje open stond, moesten zij het wel
gelooven.

Nadat hij Ambro flink de les had gelezen, zei de heer Verbrugge:

„We kunnen meneer Kaan niet dankbaar genoeg zijn en ik wil niet
wachten, hem onzen dank te gaan betuigen.”

Zoo dachten ook de ouders der andere jongens erover en meneer


Kaan ontving dien dag drie dankbare vaders, die ’t er met hem over
eens [109]waren, dat hun zoontjes schelmen van ’t zuiverste water
waren.

De heer Kaan vond het noodig nog even tot hun verdediging aan te
voeren, dat het wel rakkers waren, maar dat bij geen van het
ondeugende stel een kwaad hart zat.
[Inhoud]
HET GEHEIMZINNIGE APPARAAT.

Ambro en Karel loopen langs de Schiekade. Ze komen van


gymnastiek en daar ’t Woensdagmiddag is, willen ze nog naar het
Hol, waar ze weten dat de jongens hen wachten.

Een steenen paal, langs den weg, lokt ze echter eerst tot haasje-
over springen. De paal is hooger dan Ambro’s schouder.
Niettegenstaande dit, weet hij er met een flinken afzet handig over
heen te komen.

Karel heeft er meer moeite mee en al eenige malen heeft zijn buik
op een onaangename manier kennis gemaakt met den harden paal.

Eindelijk gelukt het hem zijn aanloop groot genoeg te nemen en hij
vliegt met een vaart over den paal heen, waarbij echter dezen keer
zijn zitvlak in zoo’n hevige botsing komt met den paal, dat dit een
eind maakt aan het spel.

Ambro staat te gieren van pret om de pijnlijke grimassen die Karel


maakt bij het wrijven langs bovengenoemd lichaamsdeel.

„Je bent ook een reuzen-os,” zegt Ambro.

„Je moet je aanloop grooter nemen.”

„Ik vertik ’t langer,” zegt Karel kwaad. [110]

„Als je nu niet meegaat, ga ik alleen.”

„Goed,” zegt Ambro. „En dan nemen we den paal mee.” Hij doet
alsof hij den paal met alle geweld uit den grond wil trekken.
„Maciste, de groote krachtpatser is er niks bij,” lacht Karel. „Schiet
op,” laat hij er op volgen, „anders komen we nooit verder.”

Het tweetal vervolgt zijn weg.

Maar daar krijgt Ambro het pontje in de gaten, dat voor twee centen
de liefhebbers naar den overkant brengt.

„Nou weet ik een goeie,” zegt hij.

„Wat dan?”

„We zullen dien pontjes-baas er tusschen nemen.”

„Hoe dan?”

„Ik loop zoo hard als ik kan om, en als ik aan den overkant ben,
wacht jij af aan welken kant het pontje ligt en dan doen we net of we
elkaar niet kennen. Ligt het nou aan mìjn kant, dan doe jij net of je
mee wil. Ik doe natuurlijk net zoo als ie aan joùw kant ligt. Vaart hij
nou over, dan smeer je ’m! In dien tusschentijd sta ik aan de andere
kant en probeer hetzelfde. Als dàt lukt, is ie volmaakt.”

„Potverdrie, ja! dàt zullen we ’m lappen,” zegt Karel.

In vollen draf rent Ambro de Heulbrug over en nu loopt het tweetal


op gelijke hoogte, ieder aan een kant van de Schie. Het pontje komt
juist, met twee juffrouwen als vracht naar Karel’s kant toe. [111]

Karel vervolgt kalm zijn weg.

Het pontje is intusschen aangekomen en Ambro staat aan den


overkant kalm en geduldig te wachten, tot de pontjesbaas hem komt
halen.
Deze keert dan ook onmiddellijk terug en zonder dat de man het in
de gaten heeft, is Ambro reeds verdwenen en houdt hij zich schuil in
het portiek van een huis.

De man begrijpt er niets van en zal wel gedacht hebben, dat hij
droomde. Nog net ziet hij een jongetje staan en—weg is ie!

„Zoo’n aap,” bromt de man. „Laat me daar voor niks overvaren! Dat
zal me geen tweeden keer gebeuren!”

Hij blijft nog even wachten en als hij eenige passagiers bij elkaar in
z’n bootje heeft, onderneemt hij de reis naar den overkant.

Dit is een groote teleurstelling voor Karel, want nu het bootje vol is,
kan hij hem den tocht niet te vergeefs laten maken. Dùs blijft hem
niets anders over, dan Ambro te fluiten en hem toe te schreeuwen
weer samen te komen op het Hofplein.

„Jammer, dat ie de tweede keer mislukte,” zegt Ambro, als ze weer


bij elkaar zijn.

„Zeg, jôh, wat is ’t al laat; ga je mee naar ’t hol?”

„Mij best, de anderen zijn er nu ook.”

Nu gaat het in looppas naar den Dierentuin. Bij het hol gekomen,
laat Ambro een zacht fluitsignaal hooren, dat onmiddellijk wordt
beantwoord en, na even goed rond gekeken te hebben, schieten zij
als konijnen weg tusschen de dichte bladeren. [112]

Op hun buik voortkruipend, komen ze eindelijk op hun zorgvuldig in


orde gehouden paadje en bevinden zich alras in het hol, waar zij
Chris en Puckie aantreffen. Beiden liggen te lezen.

„Wat zijn jullie laat,” zegt Chris.


„We hebben keet gemaakt aan de Schie.”

En ze vertellen hun maar half gelukte grap met den pontjesbaas.

Puckie vindt het „een flauwen bak”.

„Omdat jij ’m niet verzonnen hebt,” bijt Ambro hem toe.

„Neen, dan hebben wij een beteren voor vanavond,” zegt Chris. „Hè,
Puckie?”

„Nou, ’n fijne!” stemt Puckie toe.

„’t Zal wat zijn,” smaalt Ambro.

„Als jij ’t niet verzint, kan ’t niet deugen,” zegt Chris.

„Wat zijn jullie strijdlustig vandaag,” zegt Karel, die zich behagelijk
heeft neergevleid op den zachten grond en naar de plekjes blauw
van den hemel kijkt, die door de boomen zichtbaar zijn.

„A la routschkia de Baviani, met den kikvorsch op den schouder, van


de boesa van de bombernades!” zegt Ambro met een
onweerstaanbaar mal gezicht.

„Wàt zeg je nou?” vragen de jongens vol verbazing.

„Dat is Russisch,” zegt Ambro gewichtig.

„Wat beteekent het?” vraagt Chris geamuseerd.

„Juffrouw, haal je katje naar binnen, want het begint te regenen.” [113]

„Och, malle, hou je snater,” giert Puckie.

„Toe, zeg ’t nog eens,” vraagt Chris.


„Neen, ik zing me liedje geen twee keer. Kom op met je plan, Chris,
wat is ’t?”

En ze komen er mee voor den dag.

Chris neemt het woord en zegt met het gezicht van een toovenaar:

„Ik zie kans om in een heele straat belletje te trekken, door maar één
bel aan te raken.”

„Da’s onmogelijk,” zegt Karel.

„Neen, dàt kan niet,” bevestigt Ambro.

„En òf ’t mogelijk is,” zegt Chris. „Vanavond, als ’t donker genoeg is,
moeten jullie komen kijken en dan zal ik je bewijzen, dat ik waarheid
spreek.”

„Top,” zegt Ambro. „Hoe laat moeten we komen, en waar?”

„Ik heb vioolles,” zegt Karel bedrukt.

„Tot hoe laat?”

„Tot half acht.”

„Nou, dat is net mooi, want dan is ’t donker genoeg om te beginnen.”

„Zeg nou es hoe je dat doet?” vraagt Ambro nieuwsgierig.

„Je zal ’t vanavond zien,” zegt Chris op vastberaden toon. „Komen


jullie nou om kwart voor achten op de Heulbrug.”

„Wat een eind weg,” klaagt Karel.

„Ja, in een drukke straat gaat ’t niet.”


„Nou, vooruit dan. Ik zal er zijn. En we zullen Piet en Paul ook nog
even waarschuwen.”

„Ja, da’s best.” [114]

De klok slaat half acht.

Op de Heulbrug wachten Chris en Puckie en kijken in donker uit, of


ze de anderen nog niet zien verschijnen.

En jawel! in de verte hooren ze een vroolijken deun fluiten en even


daarna komen hun makkers, in marsch-tempo, aangestapt.

„Hallo!” schreeuwt Ambro. „Daar waren we.”

„Het spul kan beginnen,” lacht Chris.

Op een drafje gaat ’t nu naar de bewuste zijstraat van den Bergweg


waar het wonder vertoond zal worden.

Het is intusschen al goed donker geworden.

De straat, die Chris uitgezocht heeft tot het terrein zijner operatiën, is
wel een der stilste van deze wijk, hetgeen voornamelijk komt,
doordat ze doodloopt en begrensd wordt door een hooge houten
schutting.

Geen enkel levend wezen vertoont zich in de straat, uitgezonderd


een zwarte poes, die, door een krijgskreet van Ambro opgeschrikt, in
dwaze vlucht wegijlt.

„Hou je bakkes,” zegt Chris met ingehouden woede. „Ze mogen ons
niet hooren!”

Met „ze” bedoelde hij de argelooze bewoners der straat.


„Nou jongens,” zegt Chris. „Ik zal aan ’t laatste huis beginnen, mot je
kijken.”

Hij haalt uit z’n zak een groot kluwen stevig touw. Het einde daarvan
maakt hij vast aan den deurknop van het huis en loopt dan, het touw
[115]flink strak houdend, naar den overkant, waar hij het touw aan de
trekbel bevestigt.

Dan snijdt hij het touw af en bindt aan den deurknop van hetzelfde
huis weer een eind vast, om vervolgens de bel van het huis aan den
overkant daarmee te verbinden.

En zoo gaat hij voort, tot hij bij het zesde huis gekomen, bemerkt
geen touw meer te hebben en den arbeid dus moet staken.

„Nou, opgelet, jongens,” zegt Chris. „Nou zal je een lol beleven.”

Mèt begeeft hij zich naar het eerste huis, waar hij het touw aan den
deurknop vastbond en trekt hard aan de bel.

Het resultaat is schitterend!

In een portiek staan de jongens te kijken naar het effect.

De eerste deur wordt geopend, maar diezelfde deur brengt de bel


van den overkant in beweging. Ook daar wordt geopend en met
hetzelfde resultaat. En warempel, geen enkele deur die zijn plicht
niet doet, geen enkele bel, die weigert.

Dan verschijnen zes hoofden aan beide zijden van de straat, en


nijdige uitroepen weerklinken van „wie daar!!!” „Wie belt daar toch
zoo!!”

Een juffrouw ontdekt het touw dat aan haar deurknop is bevestigd en
in haar verwoede pogingen om het touw te verwijderen, klingelt, door
die beweging, de bel bij haar overbuurman, als bezeten door het
huis, met het gevolg, dat een man in hemdsmouwen naar buiten
komt rennen, die eveneens [116]het touw ontdekt en de verschrikte
juffrouw aan den overkant vraagt, of ze gek is geworden en wat dat
bellen van haar te beduiden heeft.

„Ik bel niet,” zegt de juffrouw nijdig.

„U belt wèl,” houdt de man vol.

„Nou kom dan kijken,” zegt de juffrouw met een hoofd als een
kalkoenschen haan.

Intusschen zijn nog meer bewoners naar buiten gekomen, die geen
van allen begrijpen hoe dit fopschellen zonder de aanwezigheid van
kwajongens kan plaats hebben.

De jongens in het portiek rollen over elkaar van het lachen en


besluiten, alvorens ontdekt te worden, het hazenpad te kiezen en in
volle vaart hollen ze naar de Heulbrug terug. Ze worden niet
achtervolgd, want de menschen hebben te veel werk om de touwen
van hun knoppen en bellen te verwijderen, maar de liefelijkste
verwenschingen worden ze achterna geroepen.

Geheel buiten adem door het harde loopen, komt het stel op de
Heulbrug aan.

„Hè, hè!” zegt Ambro voldaan. „Hij was van den bakker! Je hebt er
alle eer van, Chris! Hoe ben je op ’t idee gekomen?”

„Dat weet ik zelf niet. Maar we zullen ’t meer doen! ’t Is alleen


jammer, dat je je touw kwijt raakt bij zoo’n grap. ’t Was vast een eind
van 100 Meter.”
„Nou, dan leggen we botje bij botje en we koopen weer eens een
kluw.”

Dit vonden ze allen best. [117]

Het was intusschen kwart over acht geworden en hoog tijd om naar
huis te gaan.

Dus keerde de vroolijke troep lachend en joelend huiswaarts en


Ambro droomde dien nacht dat hij stond in een grooten kring van
juffrouwen en vloekende kerels die groote schellen in ’t rond
zwaaiden en hem steeds dichter naderden.
[Inhoud]
VAN EEN VLIEGER EN EEN MEISKE.

„Hè! is Ambro er weer niet?”

Deze uitroep verbreekt de stilte die in het Hol heerschte, waar de


jongens rustig zaten en lagen te lezen.

Piet kijkt uit z’n boek op en zegt eveneens:

„Neen, hij is er weer niet, da’s warempel al de derde keer, dat ie


ontbreekt, en na school gaat ie altijd direct naar huis. Zóó vreemd is-t-
ie nog nooit geweest.”

„Zou ie kwaad zijn om ’t een of ander?” vraagt Paul.

„Kwaad! Daar is heelemaal geen reden voor.”

Het gezicht van Chris, die al dien tijd gezwegen heeft, verraadt, dat hij
meer van de zaak weet.

„Chris, jij weet er van,” zegt Karel. „Ik zie ’t aan je heele bakkes.”

„Och, kletsika! Ik weet van niks,” bromt Chris.

„Je liegt, man! Jij weet er meer van,” houdt Karel vol.

Dan wordt Chris kwaad. [118]

„En al zou ik er meer van weten, dan hoef ik ’t jou toch niet aan je
neus te hangen.”

„Ah! zei ik ’t niet,” triompheert Karel. „Wat heeft ie je beloofd als je niks
zegt?”
„Een oplababbel als ik ’t wèl zeg,” lacht Chris.

„Wat kan ’t mij ook verder schelen,” roept Puckie, die met dezen
kwasi-onverschilligen uitroep een laatste poging doet om achter het
geheim te komen.

„Dat dacht ik ook,” zegt Chris, die zich niet in de kaart laat kijken.

Karel geeft ’t echter nog niet op.

„Och, jôh! Chris doet maar net of ie iets weet, hij weet net zooveel als
wij.”

„Zoo is ’t,” zegt Chris, maar z’n gezicht logenstraft zijn woorden.

De jongens gaan weer door met lezen, maar hun nieuwsgierigheid is


geprikkeld en lang kunnen ze er niet over zwijgen.

„Weet je wat ik doe,” zegt Puckie plotseling. „Ik ga naar z’n huis; ik zàl
weten waar ie uithangt.”

„Dat zal je laten,” zegt Chris woedend.

„Daar heb jij niks over te zeggen,” bijt Puckie hem toe.

„Heb ’t lef es te gaan!” dreigt Chris.

„Dat zal je gewaar worden,” zegt Puckie, terwijl hij opstaat.

Chris, die op den grond ligt, pakt Puckie’s been vast. Deze begint te
trappen en te schreeuwen.

„La’me los!”

„Ga dan niet naar z’n huis!” [119]

„Los!” blijft Puckie schreeuwen en geeft Chris een klap op z’n hoofd.
De andere jongens, die eerst pleizier hadden in de ruzie, zijn nu bang,
dat door al dat lawaai hun heiligdom wel eens ontdekt zou kunnen
worden. Ze trachten dus sussend tusschenbeide te komen, hetgeen
ze ook gelukt, want Chris laat Puckie’s been los en deze laatste houdt
op met schreeuwen.

Als de rust in het Hol weêrgekeerd is zegt Piet kalm:

„Ja, hoor es, Chris, als je d’r geheimen voor ons op nahoudt, dan moet
je maar je biezen pakken. We houwen voor mekaar niks achterbaks;
da’s flauwe kul! Of je zegt ’t ons, òf je gaat maar naar Ambro.”

„Natuurlijk,” stemmen de anderen toe.

Ze leggen Chris het vuur na aan de schenen en hij eindigt met Piet
gelijk te geven.

„Ja,” zegt hij dan. „Nu ie ’t mij verteld heeft, hebben jullie net zooveel
recht het te weten.”

En snel laat hij er op volgen: „Ambro heeft een meisje!”

Dat bericht valt als een bom temidden van de jongens en wordt met
gemengde gevoelens ontvangen.

„Wat een idioot!—Wat mankeert ’m!—O, is het dàt!!”—

Tot eindelijk Karel nieuwsgierig vraagt: „Wie is het?”

„Margot Hoevers.”

„Dàt schaap!—Die aanstelster!”— [120]

„En komt ie nou nooit meer met ons spelen,” vraagt Paul, die Ambro
steeds bewonderd heeft en heel dikwijls steun bij hem vond wanneer

You might also like