###E###
Asymptotic
Statistics
and mathematically
This book is an introduction
is both practical
treatment
most of
the standard topics
lihood
rank procedure
sernipararnetric
applicat
to the field of asymptotic
statistics. The
rigorous. In additi
on to
including like
of an asymptotics
course,
ficiency,
on, asymptotic ef
U-stat
s, the book also presents recent
models, the bootst
istics, and
such as
research topics
and their
rap, and empirical
processes
M-estimati
inference,
ions.
One of the unifying themes is the
experi
by limit
the local
ments. This entails
with smooth
set-up
gle, normally
of asymptotic stat
mainly
parameters
distributed
approximation
of the classical i.i.d.
approximation
by location experiments
a sin
involving
standard subjects
observation. Thus, even the
Suitable as a text for a graduate
istics are presented
in a novel
or Maste
way.
r's level
researchers
in statistics, probability,
statistics
and their
course,
applicati
this
ons
book also gives
an overview
of the latest research in asymptotic
statistics.
A.W. van
Mathemati
der Vaart is Professor of Statistics in the Department
of
cs and Computer
Science
at the Vrije Universiteit,
Amsterdam.
CAMBRIDGE S ERIES IN S TA T I S T I C AL AND PROBA B I L I S T I C
MATHEMATICS
Editorial
Board:
of Mathematics,
Utrecht University
of Statistics,
University of Oxford
Department
Department
R. Gill,
B.D. Ripley,
S. Ross, Department
M. Stein, Department
D. Willia
ms, School of Mathematical
Sciences, University of Bath
of Industrial
of Statistics,
Engineering,
University of Chicago
University of California, Berkeley
of high-qua
of stochastic applicable
vision
mathemati
textbooks
cs. The topics range from pure and applied
and expository
monographs
covers
lity upper-di
This series
all aspects
statistics to probability
theory,
gramming. The books contain
also of the state of
theoretical
made possible
operati
clear
methods,
the books also contain
the art in classical methods. While emphasizing
pro
in the field and
ions of new developments
rigorous
of
ons research, optimiza
presentat
tion, and mathematical
treatment
by advances in computational
applicat
ions and discussions
practi
ce.
of new techniques
Already
published
1 . Bootstrap Methods
Chains, by J. Norris
2. Markov
and Their Application, by A.C. Davison and D.V. Hinkley
Asymptotic
Statistics
A.W. VANDER VAART
CAMBRIDGE
UNIVERSITY PRESS
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
CAMBRIDGE UNIVERSITY PRESS
Street, Cambridge,
The Pitt Building,
United Kingdom
Trumpington
Cambridge CB2 2RU, UK http://www.cup.cam.ac.uk
USA http://www.cup.org
The Edinburgh Building,
40 West 20th Street, New York, NY 10011-4211,
10 Stamford Road, Oakleigh,
Ruiz de Alarcon 13, 28014 Madrid, Spain
Melbourne 3166, Australia
© Cambridge University
Press 1998
This book is in copyright.
to the provisions
no reproduction
the written permission
Subject to statutory
of relevant collective
licensing
of any part may take place without
of Cambridge University
Press.
exception
and
agreements,
First published
First paperback edition 2000
1998
Printed in the United States
of America
Typeset in Times Roman 10112.5 pt in LKfff(2 [TB]
A catalog
for this book is available
record
Library of Congress Cataloging
data
I A.W. van der Vaart.
Vaart, A.W. van der
Asymtotic statistics
in Publication
from the British
Library
p. em. - (Cambridge series in statistical
and probablistic
references.
mathematics)
Includes bibliographical
1. Mathematical
statistics
II. Series: cambridge series on statistical
mathematics.
CA276.V22 1998
519.5-dc21
- Asymptotic
theory. I. Title.
and probablistic
98-15176
ISBN 0 521 49603 9 hardback
ISBN 0 521 78450 6 paperback
To Maryse and Marianne
Contents
Preface
Notation
page xiii
page XV
1 . Introduction
ons
of Moments
Determining
in Total Variation
Problems
o and 0 Symbols
stic Functions
Classes
Logarithm
Theorem
Statistical
Optimality
1 . 1 . Approximate
1 .2. Asymptotic
1 .3. Limitati
1.4. The Index n
2. Stochastic
Convergence
1
Procedures 1
2
Theory
3
4
5
5
2.1 . Basic Theory
2.2. Stochastic
12
*2.3. Characteri
13
*2.4. Almost -Sure Representations 17
17
*2.5. Convergence
*2.6. Convergence-
1 8
*2.7. Law of the Iterated
19
20
*2.8. Linde
berg-Feller
22
*2.9. Convergence
24
25
25
30
3 1
32
33
34
35
35
37
40
41
41
44
5 1
3.1 . Basic Result
3.2. Variance-Stabilizing
*3.3. Higher-Order
Expansions
*3.4. Uniform Delta
Method
*3.5. Moments
5.1. Introduction
5.2. Consist
ency
5.3. Asymptotic
of Moments
4.1 . Method
*4.2. Exponential
Families
Problems
Estimators
Problems
5. M-and Z-Estimators
Normality
Vll
3. Delta
Method
Transformati
ons
4. Moment
Vlll
Contents
ood Estimators
Parameters
Likelih
*5.4. Estimated
5.5. Maximum
*5.6. Classica
l Conditions
*5.7. One-Step Estimators
*5.8. Rates of Convergence
*5.9. Argmax
Theorem
Problems
6. Contiguity
6.1 . Likelihood
6.2. Contiguity
Ratios
Problems
60
61
67
7 1
75
79
83
85
85
87
91
92
92
93
Experiment 97
7. Local Asymptotic
Normality
7.1 . Introduction
hood
7.2. Expanding
7.3. Convergence
7.4. Maximum
*7.5. Limit Distributi
*7.6. Local Asymptotic
to a Normal
the Likeli
Likelihood
Problems
8. Efficiency
of Estimators
Concent
ration
Normality
100
ons under Alternatives 103
103
106
108
108
1 10
Efficiency
8.1 . Asymptotic
8.2. Relative
8.3. Lower Bound for Experiments 1 1 1
8.4. Estimating
Means
Normal
8.5. Convolution
Theorem
8.6. Almost-Everywhere
Convolution
1 12
1 15
Theorem
*8.7. Local Asymptotic
Minimax
*8.8. Shrinkage
Estimators
*8.9. Achieving
the Bound
*8. 10. Large Deviations
Problems
of Experiments
9.1 . Introduction
9.2. Asymptotic
9.3. Asymptotic
9.4. Uniform
Distributi
9.5. Pareto
9.6. Asymptotic Mixed
9.7. Heuristics
1 15
Theorem 1 17
1 19
120
122
123
125
125
Theorem 126
127
129
130
1 3 1
136
137
138
138
140
10. 1 . Introduction
10.2. Bernstein
Representation
Normality
-von Mises Theorem
Distribution
Normality
Problems
on
9. Limits
10. Bayes Procedures