(Download PDF) Statistical Modelling Using Local Gaussian Approximation 1St Edition Tjostheim Ebook Online Full Chapter

Statistical Modelling using Local
Gaussian Approximation 1st Edition

Tjostheim
Visit to download the full and correct content document:
https://ebookmeta.com/product/statistical-modelling-using-local-gaussian-approximati
on-1st-edition-tjostheim/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
From Statistical Physics to Data-Driven Modelling

Simona Cocco
https://ebookmeta.com/product/from-statistical-physics-to-data-
driven-modelling-simona-cocco/
Heat Transfer Modelling Using COMSOL 1st Edition Layla

S Mayboudi
https://ebookmeta.com/product/heat-transfer-modelling-using-
comsol-1st-edition-layla-s-mayboudi/
Statistical Data Analysis Using SAS Intermediate

Statistical Methods Springer Texts in Statistics
Marasinghe
https://ebookmeta.com/product/statistical-data-analysis-using-
sas-intermediate-statistical-methods-springer-texts-in-
statistics-marasinghe/
Asymptotic Statistical Inference A Basic Course Using R

Shailaja Deshmukh
https://ebookmeta.com/product/asymptotic-statistical-inference-a-
basic-course-using-r-shailaja-deshmukh/
Mathematical Modelling of Decision Problems Using the
SIMUS Method for Complex Scenarios 1st Edition Nolberto
Munier
https://ebookmeta.com/product/mathematical-modelling-of-decision-
problems-using-the-simus-method-for-complex-scenarios-1st-
edition-nolberto-munier/
Gaussian Measures in Finite and Infinite Dimensions 1st

Edition Daniel W. Stroock
https://ebookmeta.com/product/gaussian-measures-in-finite-and-
infinite-dimensions-1st-edition-daniel-w-stroock/
Computation and Approximation 1st Edition Vijay Gupta
https://ebookmeta.com/product/computation-and-approximation-1st-
edition-vijay-gupta/
Marketing Analytics Statistical Tools for Marketing and

Consumer Behaviour using SPSS 1st Edition Carvalho De
Mesquita
https://ebookmeta.com/product/marketing-analytics-statistical-
tools-for-marketing-and-consumer-behaviour-using-spss-1st-
edition-carvalho-de-mesquita/
The Psychosocial Reality of Digital Travel Being in

Virtual Places Ingvar Tjostheim
https://ebookmeta.com/product/the-psychosocial-reality-of-
digital-travel-being-in-virtual-places-ingvar-tjostheim/
STATISTICAL
MODELING USING
LOCAL GAUSSIAN
APPROXIMATION
This page intentionally left blank
STATISTICAL
MODELING USING
LOCAL GAUSSIAN
APPROXIMATION
DAG TJØSTHEIM
HÅKON OTNEIM
BÅRD STØVE
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2022 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our arrangements
with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency,
can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described herein. In
using such information or methods they should be mindful of their own safety and the safety of
others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.
Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library
ISBN: 978-0-12-815861-6
For information on all Academic Press publications

visit our website at https://www.elsevier.com/books-and-journals
Publisher: Candice Janco

Acquisitions Editor: Brian Romer
Editorial Project Manager: Susan Ikeda
Production Project Manager: Surya Narayanan Jayachandran
Designer: Matthew Limbert
Typeset by VTeX
To our wives: Marit, Karina, Ellen Johanne
Contents
Biography xi
Preface xiii
1. Introduction 1
1.1. Computer code 6
References 6
2. Parametric, nonparametric, locally parametric 7

2.1. Introduction 7
2.2. Parametric density models 9
2.3. Parametric regression models 17
2.4. Time series 20
2.5. Nonparametric density estimation 23
2.6. Nonparametric regression estimation 29
2.7. Fighting the curse of dimensionality 33
2.8. Quantile regression 37
2.9. Semiparametric models 38
2.10. Locally parametric 40
References 43
3. Dependence 49
3.1. Introduction 49
3.2. Weaknesses of Pearson’s ρ 52
3.3. The copula 56
3.4. Global dependence functionals and tests of independence 61
3.5. Test functionals generated by local dependence relationships 80
References 81
4. Local Gaussian correlation and dependence 87

4.2. Local dependence 90
4.3. Local Gaussian correlation 94
4.4. Limit theorems 99
4.5. Properties 105
4.6. Examples 112
4.7. Transforming the marginals: Normalized local correlation 115
4.8. Some practical considerations 120
4.9. The p-dimensional case 123
vii
viii Contents
4.10. Proof of asymptotic results 123

References 133
5. Local Gaussian correlation and the copula 135

5.2. Local Gaussian correlation for copula models 136
5.3. Examples 142
5.4. Recognizing copulas by goodness-of-fit 148
5.5. A real-data study 157
References 159
6. Applications in finance 161

6.2. Conditional correlation and the bias problem 164
6.3. Empirical analysis of dependence of financial returns 167
6.4. The portfolio allocation problem 182
6.5. Financial contagion 196
References 209
7. Measuring dependence and testing for independence 213

7.2. Testing of independence in iid pairs of variables using local correlation
functionals 214
7.3. Testing for serial independence in time series 224
7.4. Describing nonlinear dependence and tests of independence for two
time series 235
7.5. Proofs 253
References 259
8. Time series dependence and spectral analysis 261

8.2. Local Gaussian spectral densities 265
8.3. Visualizations and interpretations 280
References 297
9. Multivariate density estimation 301

9.2. Description of the estimator 304
9.3. Asymptotic theory 308
9.4. Bandwidth selection 313
9.5. An example 316
9.6. Investigating performance in the multivariate case 318
9.7. A more flexible version of the LGDE 323
Contents ix
9.8. Proofs 327

References 333
10. Conditional density estimation 335

10.2. Estimating the conditional density 337
10.3. Asymptotic theory for dependent data 339
10.4. Examples 342
10.5. Proof of theorems 348
References 352
11. The local Gaussian partial correlation 353

11.2. The local Gaussian partial correlation 354
11.3. Properties 358
11.4. Estimation of the LGPC by local likelihood 360
11.5. Asymptotic theory 362
11.6. Examples 365
11.7. Testing for conditional independence 370
11.8. The multivariate LGPC 376
References 382
12. Regression and conditional regression quantiles 385

12.2. Comparison with additive regression modeling 387
12.3. Local Gaussian regression estimation 388
12.4. Asymptotic normality 390
12.5. Example 394
12.6. Conditional quantiles 396
12.7. Proof 398
References 401
13. A local Gaussian Fisher discriminant 403

13.2. A local Gaussian Fisher discriminant 408
13.3. Some asymptotics of Bayes risk 413
13.4. Choice of bandwidth 417
13.5. Illustrations 420
13.6. Summary remark 425
References 426
Author index 429

Subject index 437
Biography
Dag Tjøstheim
is Emeritus Professor, Department of Mathematics, University of Bergen.
He has a PhD in applied mathematics from Princeton University (1974).
He has authored more than 120 papers in international journals. He is a
member of the Norwegian Academy of Sciences and has received several
prizes for his scientific work. His main interests are in econometrics, non-
linear time series, nonparametric methods, modeling of dependence, spatial
variables, and fishery statistics.
Håkon Otneim
is Associate Professor at the Norwegian School of Economics. He has a
PhD in statistics from the University of Bergen (2016), and he has published
papers in international journals about multivariate density estimation and
conditional density estimation. His research interests include development
and application of nonparametric and semiparametric statistics, statistical
programming, and data visualization.
Bård Støve
is Professor of Statistics at the University of Bergen. He received his PhD
degree in statistics, 2005. He was Assistant Professor at the Norwegian
School of Economics (2007–2011), and worked as an Actuary in a con-
sulting firm (2005–2007). He has been working on the development of
nonparametric models and application of such models to finance and eco-
nomics. He has published several research papers in such journals as Econo-
metric Theory and Scandinavian Journal of Statistics.
xi
Preface
The central idea of this book is the approximation of a general multivariate

density f by a family of Gaussian distributions. Locally around a point x in
the support of f , f is approximated by a multivariate Gaussian distribution.
This makes it possible to define a local mean, a local variance, and a local
correlation matrix.
This idea is powerful and can be applied to a number of tasks and
problems for continuous stochastic variables. This has been done in sev-
eral recent papers. These papers, thirteen altogether, form the basis of this
book. In particular, following two introductory chapters, each of the main
Chapters 3–11 and 13 is composed from one or more of these papers.
In Chapter 4 the emphasis is on the local correlation, its properties, and
its use as a measure of dependence. It can be defined on the original x-scale,
but also on a normalized z-scale obtained by transforming the marginals of
f to the standard normal. This is in a way analogous to the transformation
to uniform variables in a copula construction, and the relationship between
the copula concept and the Gaussian approximation concept is explored in
Chapter 5.
It has long been realized that a multivariate Gaussian distribution fitted
to data in finance or econometrics may not be a good idea. In fact, data
in finance and econometrics have thick tails, and a global Gaussian fit may
lead to disastrous results with a large underestimation of economic risk. In
Chapter 6, we apply local Gaussian approximation to financial data, includ-
ing financial contagion and a preliminary attempt of portfolio construction.
With the establishment of local Gaussian correlation as a measure of lo-
cal dependence, an obvious next step is employing this measure to testing of
independence. This is done in Chapter 7 in three stages: testing of indepen-
dence between two sequences, each consisting of independent identically
distributed variables, testing of serial independence in a time series, and
testing of independence between two stationary time series. This implies
the introduction of a local autocorrelation and cross-correlation concept.
The locally Gaussian autocorrelation introduced in Chapter 7 is used in
Chapter 8 to construct a locally Gaussian spectral density. It coincides with
the ordinary power spectral density in the Gaussian time series case. For
non-Gaussian data, the new local spectral concept can be used to pick up
spectral peaks that may be hidden in a conventional spectral estimation.
xiii
xiv Preface
Chapters 9 and 10 are devoted to estimation of multivariate density

functions and to estimation of multivariate conditional densities. This is
done by merging the Gaussian densities of the local approximating families.
In the conditional case the unique properties of the conditional (global)
Gaussian distribution is of crucial importance. Further, the curse of dimen-
sionality is sought circumvented by a simplified Gaussian approximation,
in a sense similar to the use of the additive simplification in nonparametric
regression.
The local Gaussian approximation also makes it possible to introduce a
local Gaussian partial correlation. In Chapter 11, it is shown how this can
be used to construct a local measure of conditional dependence and to test
for conditional independence. We believe that this idea has potential for
network theory and causality.
Perhaps the most important use of nonparametric estimation methods
is currently in nonparametric regression. The local Gaussian approxima-
tion concept is developed for multivariate statistical analysis, where all of
the variables are, so to speak, on the same basis, in contradistinction to re-
gression analysis, where one variable or a group of variables are dependent
variables expressed as a function of another group of explanatory variables.
Nevertheless, in Chapter 12, we make an attempt to apply local Gaussian
approximation techniques to regression estimation and quantile regression
estimation. More work is required to determine in what way local meth-
ods can complement nonparametric techniques like, for instance, additive
modeling.
The local Gaussian approach can be applied to other fields of statistics
as well. As an example of such an application, in Chapter 13, we look
at applications to classification and discrimination, involving among other
things a local Fisher discriminant.
To put local Gaussian approximation analysis into context with other
methods, the book also contains three introductory chapters. Chapter 1
contains a general introduction explaining the overall features and con-
cepts of our approach. Chapter 2 gives a brief but at the same time quite
broad overview over parametric, nonparametric, and locally parametric ap-
proaches to statistics. Chapter 3, based on a very recent survey paper to
appear in Statistical Science, contains an overview of the statistical concept
of dependence, how it can be measured, and how we can test for inde-
pendence. The survey concentrates on methods developed in the last two
decades, going beyond the most used measure of dependence, the Pearson
Preface xv
correlation. The local Gaussian correlation is put directly into this context
in Chapter 4.
There is some overlap between the various chapters in the book. This
has been done intentionally, so that a reader can single out the chapters of
primary interest to her/him. Most chapters can be read independently of
each other as the basic material from Chapter 4 is included briefly as an
introductory material in each of the following chapters. The mathematical
and technical level of each chapter is quite modest. For readers with more
interest in technical details, we give references, often to supplementary ma-
terial to the papers that the book is composed from.
There are three R-packages that have been developed for various types
of analysis in the book. We do not present details of use of these packages
in this book, but references to the packages are given in Chapter 1.
The local Gaussian approach is a recently developed methodology. Some
of the chapters are based on papers that have just appeared or are in the
process of appearing in journals. Putting this in a book, the emphasis is on
presenting the fundamental concepts inherent in a local Gaussian approx-
imation and in demonstrating their usefulness in several areas in statistics.
At the same time, we hope that the book may serve as a starting point and
inspiration for further research and applications in the subject matters taken
up in each of the chapters of the book, as well as in new subject areas.
The chapters of the book have been primarily based on papers by the
three authors of the book, but some chapters have also benefited from joint
work and joint papers with others, namely Karl Ove Hufthammer (Chap-
ters 4 and 6), Geir Berentsen (Chapters 5 and 7), Viginia Lacal (Chapter 7),
Lars Arne Jordanger (Chapter 8), Martin Jullum (Chapter 13), and Anders
Sleire (parts of Chapter 6). Without their contributions the book had not
been possible in its present form, and we are very grateful to them for their
good work and cooperation on these subjects.
Dag Tjøstheim
Håkon Otneim
Bård Støve
Bergen, May 2021
CHAPTER 1
Introduction
Contents
1.1. Computer code 6
References 6
The most important distribution in statistics is the Gaussian distribu-

tion. It has a number of very useful and special properties, particularly
in the multivariate case. Just think about a normally distributed vector
X = (X1 , . . . , Xp )T of dimension p, where T denotes transposed. Its dis-
tribution is given by the density function
1
f (x) = exp{(x − μ)T −1 (x − μ)},
(2π)p/2 ||1/2
where μ = {μi } and = {σij }, i, j = 1, . . . , p, are the mean vector and co-
variance matrix of X, respectively. Looking at this familiar expression, it
is easy to forget its simplicity and elegance. Here we have a distribution
whose location is completely determined by its means μi , the scale by the
variances σii , and whose dependence relations have the amazing property
that they are completely determined by the pairwise covariances σij .
Moreover, if X is subdivided into two components X = (X 1 , X 2 ), then
any linear combination of X 1 and X 2 is again Gaussian, and the condi-
tional distribution fX 1 |X 2 (x1 |x2 ) is Gaussian. The dependence properties of
these derived distributions are again determined by the pairwise covari-
ances, in the latter case, through the partial covariances. These properties
make the Gaussian especially suitable for linear statistical modeling. Fur-
ther, the properties of the conditional distribution imply that in a Gaussian
system the optimal least squares predictor, given by the conditional mean,
is linear and equals the optimal linear predictor. Finally, uncorrelatedness is
equivalent to independence in the Gaussian distribution, that is, X 1 and X 2
are independent if and only if they are uncorrelated. In this case, we can
test for independence by computing covariances.
Unfortunately, data are not always well described by a Gaussian distri-
bution and a linear model. In particular, for data in economics and finance
the data are usually governed by distributions having thicker tails, and the
dependence properties are not well described by pairwise covariances only,
Statistical Modeling using Local Gaussian Approximation Copyright © 2022 Elsevier Inc.
https://doi.org/10.1016/B978-0-12-815861-6.00008-0 All rights reserved. 1
2 Statistical Modeling using Local Gaussian Approximation
as is inherent in the Gaussian distribution. In fact, assuming a linear Gaus-

sian model can lead to disastrous results with drastic underestimation of the
risk involved in economic and financial transactions; see, for example, Taleb
(2007).
To some degree, these problems can be avoided, or at least lessened, by
trying to fit other parametric families to the data or by using a semiparamet-
ric or nonparametric approach. An important concept in a nonparametric
methodology is the concept of a local approximation. In nonparametric
density estimation, we may take as the starting point a locally smoothed
version of a histogram of the available data. In a regression, the regression
relationship may be approximated by locally fitted polynomials (the partic-
ular case of the locally constant case being the regression kernel estimator).
What is local is determined by a bandwidth parameter, which for a given
point x, selects the neighboring points close to x, “close” being determined
by the bandwidth acting as a distance measure. A density, a conditional den-
sity, or a regression can be estimated nonparametrically in this manner. As
the dimension p increases, the curse of dimensionality emerges, and sim-
plifying assumptions, such as the additive model for regression, have to be
introduced.
We are now ready to formulate the main idea of this book and of
the papers it consists of. The idea is simply to approximate an arbitrary
p-dimensional density function f locally by a family of Gaussians distribu-
tions. This can be viewed as an example of a semiparametric or a locally
parametric approach. In principle, another family of parametric distribu-
tions could be used as a local approximant (as has been done by Hjort and
Jones (1996), who considered the locally parametric density estimator), but
we believe that the well-known simple and elegant properties of Gaussians
makes this family of distributions the optimal choice. So, for a point x, in
a neighborhood of x, we fit a Gaussian distribution, the neighborhood be-
ing determined by a bandwidth parameter. The parameters of this Gaussian
distribution will be functions of the coordinates of the point x. Moving
to another point y and fitting another Gaussian in the neighborhood of
y will in general result in another set of parameters depending on y. The
exception is when f itself is Gaussian. In that case, as the number of avail-
able observations tends to infinity at the same time as the bandwidth tends
to zero, the estimated parameters at x and y will ultimately coincide and
be equal to the parameters of the Gaussian f . The advantage of using the
Gaussian distribution as an approximating family is that the unique proper-
ties of the Gaussian can be locally used for a general, possibly non-Gaussian,
Introduction 3
density f . For instance, for a thick-tailed distribution, it can be locally ap-

proximated in the tail by a Gaussian with large variance. In the multivariate
case, we have the potential of approximating multivariate tail behavior lo-
cally by an appropriate multivariate Gaussian. This turns out to be useful
for multivariate financial market data.
Using this idea as a statistical modeling philosophy has many ramifi-
cations and applications as we try to illustrate throughout the book. For
example, we can define local covariances and correlations, and even lo-
cal partial covariances. Local dependence and conditional dependence can
be measured by these quantities, and local independence and conditional
independence can be tested. Dependence properties may be analyzed by ag-
gregation, and independence may be tested over larger regions, ultimately
over the entire range of the data. Further, by using the same principle in
approximating locally the joint distribution of time series variables, we can
introduce concepts of local autocorrelation and cross-correlation. A local
spectral density can be constructed, which makes it possible to derive one
local spectral density describing the oscillatory properties at one level (e.g.,
close to extremes) and another local one describing the frequency distribu-
tion close to the center of the data.
Seen from this perspective, the unique properties of the Gaussian can be
utilized in a non-Gaussian environment but, again, locally. In this book, we
discuss many such examples, but there are other avenues that have not been
explored so far. Local principal component analysis is one of them. The
ordinary principal components are found by solving an eigenvalue problem
involving the covariance matrix. Local principal components can be found
by replacing the ordinary global covariance matrix by a matrix of local
covariances. Another area where research has been initiated, is multiple
spectral analysis, where a local amplitude spectrum and phase spectrum can
be introduced. There are other possibilities as well, and we believe local
Gaussian approximation to be a very comprehensive tool.
Another potential extension is to broaden the Gaussian family to a more
general family like the family of elliptic distributions. Then we lose some of
the simple properties of the Gaussian (e.g., independence is not equivalent
to uncorrelatedness), but, on the other hand, a multivariate t-distribution,
belonging to the elliptic family, is much easier to approximate. Other fam-
ilies are considered by Hjort and Jones (1996), but just in the situation of
deriving an alternative density estimator to the kernel estimator.
Local quantities like the local correlation is sensitive to the curse of di-
mensionality as the dimension of X increases. Throughout the book, we
discuss ways of bypassing it. Somewhat similarly to the additive approxi-

mation in regression analysis, we try to approximate the elements σij (x) of
the covariance matrix by a function of two coordinates σij (xi , xj ). Still we
have to be careful at the edges of the data set where there are few observa-
tions. Another device that has been much used in this book is transforming
the data to a standard normal marginal scale by using the marginal empir-
ical distribution function. A very different approach, which may deserve a
closer examination, is trying to fit a parametric model to the local quanti-
ties.
Here is a brief overview of the contents of the book:
To put our method into perspective, we summarize briefly traditional
parametric, semiparametric, nonparametric, and locally parametric mod-
eling in Chapter 2. Very briefly, properties of the Gaussian and elliptic
distribution are also included.
Local correlation represents one way of measuring dependence as a lo-
cal version of the traditional Pearson correlation. Chapter 3 presents a fairly
self-contained review of recent developments in nonlinear dependence
analysis, among them, the Brownian distance covariance and reproducing
kernel Hilbert space measures, both having received considerable attention
lately.
In Chapter 4, we contrast the dependence measures of Chapter 3 with
the local Gaussian correlation (LGC). We define this concept as well as the
concept of local Gaussian approximation. Two versions are described, one
on the original x-scale and one on the z-scale obtained by transforming
the marginals to standard normals. We give a number of properties and
illustrate these on simulated and financial real data.
The connection to description of dependence by means of copulas is
explored in Chapter 5. We compute the local correlation for traditional
copulas like the Clayton, Gumbel, and Frank copulas.
Much of what we do is motivated by problems met in the description of
financial markets. Chapter 6 contains a number of applications to financial
and econometric data. It is shown that key “typical” dependence properties
of such markets are well described by the local Gaussian correlation, and
we also look at applications to financial contagion, portfolio analysis, and
value at risk.
In Gaussian distributions, independence and uncorrelatedness are
equivalent, so it is natural to use the local Gaussian correlation to test
for local and nonlinear independence. In Chapter 7, we extend this to tests
of serial dependence in a univariate time series and to independence test-
Introduction 5
ing between two time series. We compare to other tests like the Brownian
distance covariance for both simulated and real data.
In Chapter 8 the time series frame is kept, but here we focus on the
local autocorrelation and the local spectrum that can be derived from it.
It is shown that frequency behavior that cannot be detected by ordinary
spectral analysis can be detected by the local spectrum. The chapter also
contains a brief review of alternative nonlinear spectral techniques.
Chapters 9 and 10 are devoted to density estimation and conditional
density estimation, respectively. The density estimation is the aspect stressed
by Hjort and Jones (1996) in their local parametric analysis. We carry this
through for the local Gaussian approximation of a density and compare
with other methods as the dimension increases. In the conditional density
estimation, we exploit locally the fact that the conditional density in a joint
Gaussian density framework is again a Gaussian density, where the local
mean vector and covariance matrix can be found by explicit formulas.
In a sense testing for conditional independence is more important than
testing for independence. This is due to the applications to causality analysis
among other things. For globally Gaussian data, the partial correlation co-
efficient is an important tool, for example, in path analysis. In Chapter 11,
we introduce the local partial correlation and use it both for measur-
ing conditional dependence and for testing of conditional independence.
We compare with alternative tests and give applications to testing Granger
causality.
Regression and conditional quantile estimation is covered in Chap-
ter 12. We note that the local Gaussian approach is primarily suited to a
situation where all the variables are treated on the same basis. It is perhaps
less well suited to a situation where there is one dependent variable and
one or several explanatory variables. Nevertheless, we show in this chapter
that the local Gaussian approximation can be applied and that in particular
cases it may offer an alternative to the additive approximation in regression
models.
The traditional Fisher discriminant for discriminating between two or
more populations is based on a Gaussian assumption. In Chapter 13, we
make the parameters of the Gaussian local and derive a local Gaussian Fisher
discriminant, which is applied to simulated and real data. It is easy to find
examples where the global Fisher discriminant does not work, whereas the
local one does.
1.1 Computer code

The package lg, see Otneim (2021), for the R programming language (see
R Core Team, 2017) provides implementations of most of the methodolog-
ical advances on applications of the local Gaussian approximation presented
in this book. This includes estimation of the local Gaussian correlation it-
self, multivariate density estimation, conditional density estimation, various
tests for independence, conditional independence and financial contagion
(cf. Chapter 6), and a graphical module for creating dependence maps;
see Otneim (2019). Note that the use of local Gaussian correlation in
spectral analysis of time series, presented in Chapter 8, has its own com-
putational ecosystem in the localgaussSpec-package1 for R. The R package
localgauss (see Berentsen et al., 2014) provided the first publicly available
implementation of the LGC and a test for independence. Note that the
lg-package depends on the localgauss-package.
We refer to the R documentation of the mentioned packages and Ot-
neim (2021) for the direct use of various available functions, as this will not
be covered in the book.
References
Berentsen, G.D., Kleppe, T., Tjøstheim, D., 2014. Introducing localgauss, an R package for
estimating and visualizing local Gaussian correlation. Journal of Statistical Software 56
(12), 1–18.
Hjort, N., Jones, M., 1996. Locally parametric nonparametric density estimation. Annals of
Statistics 24 (4), 1619–1647.
Otneim, H., 2019. lg: Locally Gaussian distributions: estimation and methods. https://
CRAN.R-project.org/package=lg. R package version 0.4.1.
Otneim, H., 2021. Ig: an R package for local Gaussian approximations. To appear, The R
Journal. URL: https://journal.r-project.org/archive/2021/RJ-2021-079/index.html.
R Core Team, 2017. R: A Language and Environment for Statistical Computing. R Foun-
dation for Statistical Computing, Vienna, Austria.
Taleb, N.N., 2007. The Black Swan: The Impact of the Highly Improbable. Random
House.
1 See https://github.com/LAJordanger/localgaussSpec for details.

CHAPTER 2
Parametric, nonparametric,
locally parametric
Contents
2.1. Introduction 7
2.2. Parametric density models 9
2.2.1 The Gaussian distribution 9
2.2.2 The elliptical distribution 11
2.2.3 The exponential family 15
2.3. Parametric regression models 17
2.3.1 Linear regression 17
2.3.2 Nonlinear regression and some further modeling aspects 19
2.4. Time series 20
2.5. Nonparametric density estimation 23
2.5.1 Nonparametric kernel density estimation 24
2.5.2 Bandwidth selection 26
2.5.3 Multivariate and conditional density estimation 27
2.6. Nonparametric regression estimation 29
2.6.1 Kernel regression estimation 29
2.6.2 Local polynomial estimation 31
2.6.3 Choice of bandwidth in regression 32
2.7. Fighting the curse of dimensionality 33
2.7.1 Additive models 34
2.7.2 Regression trees, splines, and MARS 37
2.8. Quantile regression 37
2.9. Semiparametric models 38
2.9.1 Partially linear models 39
2.9.2 Index models and projection pursuit 39
2.10. Locally parametric 40
References 43
2.1 Introduction
In statistical modeling, we have to choose between a parametric model
and the use of nonparametric statistics. A compromise is a semiparametric
model, where both aspects of modeling are taken into consideration.
For a parametric model, the mathematical form of the model and
relationships between stochastic variables entering the model and their dis-
tributions are explicitly stated and generally assumed to be known except
Statistical Modeling using Local Gaussian Approximation Copyright © 2022 Elsevier Inc.
https://doi.org/10.1016/B978-0-12-815861-6.00009-2 All rights reserved. 7
for a set of parameters. The parameters may, for instance, appear as coeffi-
cients in a linear regression or as parameters of a distribution function from
a certain class. Strictly speaking, a parametric model is never true, or in the
words of Box and Draper (1987, p. 424), “All models are wrong, but some
are useful.” A parametric model is often quite simple and has a straightfor-
ward interpretation. In certain situations, however, parametric models may
lead astray. When the model is seriously wrong, the accuracy of param-
eter estimates does not help. One well-known example is the estimation
of value-at-risk for financial markets. Sometimes, financial crises have been
blamed on using Gaussian models in the tail of a distribution when financial
objects quite clearly have thicker tails. Using the Gaussian distribution may
lead to a disastrous underestimation of the risk; see Taleb (2007). In such
situations a parametric model works as an extremely inconvenient strait
jacket.
The purpose of a nonparametric approach is letting the data speak for
themselves and thereby preventing such situations from occurring. How-
ever, this approach also has its disadvantages. First, the convergence rate of
nonparametric estimates is slower than the parametric rate. But perhaps the
most important obstacle is the curse of dimensionality: when we have a
moderate or large number of variables, the nonparametric approach does
not work in practice. It may still be possible to state theorems of conver-
gence rates for nonparametric estimates, but these are so slow that we need
astronomically large sample sizes to come close to the true values, which
we typically do not have. There are various ways of trying to get around
the curse. This can be done by assuming further restrictive assumptions such
as an additive model in a regression context. We will come back to this on
several occasions later in this chapter and in later chapters of the book.
Another way of tackling the curse of dimensionality is using a semipara-
metric model, that is, a model where some parts of the model are treated
parametrically and other parts of the model are treated nonparametrically.
The implicit understanding is that the nonparametric part is specified in
such a way that we avoid the curse of dimensionality. However, it may not
be obvious which parts of the model should be specified parametrically and
which parts should be treated in a nonparametric fashion.
The main philosophy of this book is to try to take advantage of the best
features of the nonparametric and parametric methodologies. We do this
by letting the parameters of a parametric model depend on the variables
involved, which is a local parametric approach advocated by, for example,
Hjort and Jones (1996) and Loader (1996). These two references treat lo-
Parametric, nonparametric, locally parametric 9
calization of parameters for a quite general parametric family. On the other

hand, we concentrate on localizing the Gaussian parametric family. It turns
out that this has some great potential advantages, and we are not only able
to use the local Gaussian structure to lessen the curse of dimensionality, but
we are also able to generalize and extend models where the correlation or
autocorrelation function plays a natural role. A brief introduction to the
idea of local parametric estimation is given in Section 2.10 of this chapter,
and a much more detailed account of the local Gaussian approach follows
in Chapter 4.
In the present chapter we will give a brief overview of the two main
approaches, parametric and nonparametric, and mention the semiparamet-
ric hybrid. The overview will be rather subjective in its choice of topics
and emphasis. Some special topics are given more coverage than in most
reviews of this sort. We do this in such a manner because we not only mo-
tivate local Gaussian modeling in its present state, but also briefly include
topics that may serve as extensions of the local modeling as presented in
this book. There are no data illustrations in this chapter as such illustrations
can be found in numerous other books on these topics.
2.2 Parametric density models

Motivated by their use later in the book, as possible candidates for exten-
sions of local Gaussian models, we discuss three types of distributions: the
Gaussian, elliptic, and exponential families.
2.2.1 The Gaussian distribution

The Gaussian, or normal, distribution is probably the most used distribu-
tion in statistics. There are several reasons for this, but the most important
one is probably the central limit theorem, which states that averages are
(approximately) normally distributed if we have fairly many observations.
Moreover, the normal distribution has a number of attractive and simple
mathematical properties, not the least that it can be generalized very easily
to the multivariate case. This is one reason why many books on multivariate
statistics to a large part are based on the multivariate normal distribution;
see, for example, Anderson (2003) and Johnson and Wichern (2007).
In applications, however, and perhaps particularly in finance, the nor-
mal distribution does not always give a good approximation to the data and
may in fact lead to very wrong and even catastrophic results. The main idea
of this book is to use the Gaussian distribution as a local approximation.
We are seeking to replace a given distribution, univariate or multivariate,

by a family of Gaussian distributions, where each member of the family
approximates the properties of the given distribution locally in a neighbor-
hood surrounding a given point. The motivation is that we hope to avoid
the inaccuracy of using the pure Gaussian distribution and at the same time
retain simple and unique properties of the Gaussian distribution locally.
The Gaussian is of course a prime example of a parametric model in
statistics, and in this section, we very briefly survey some fundamental
facts of this distribution that we will use in later chapters. The multi-
variate normal density function f (x) for a continuous stochastic variable
X = (X1 , . . . , Xp )T of dimension p is given by

f (x) = (2π)−p/2 ||−1/2 exp (x − μ)T −1 (x − μ) , (2.1)
where μ = (μ1 , . . . , μp )T with E(Xi ) = μi , and is the covariance matrix

of X given by = {σij } with σij = Cov(Xi , Xj ) = E (Xi − μi )(Xj − μj ) for
i = j and σii = σi2 , the variance of Xi . Moreover, || is the determinant of ,
and T denotes transpose. We write X ∼ N (μ, ). A fantastic property of
this multivariate distribution is that it is composed of pairwise dependencies
only. It suffices to compute the means, variances, and pairwise covariances
to obtain the whole distribution. This is a fact that we will use later in our
local Gaussian analysis.
In the bivariate case (p = 2), we write
1 1
f (x) = exp − ×
2πσ1 σ2 1 − ρ12
2 2(1 − ρ12
2 )
x −μ 2 x1 − μ1 x2 − μ2 x2 − μ2 2
1 1
− 2ρ12 + ,
σ1 σ1 σ2 σ2
where ρ12 = σ12 /σ1 σ2 is the correlation between X1 and X2 . It is well
known that uncorrelatedness does not imply independence in general, but
for the Gaussian, as is easy to check from the form of the distribution func-
tion, uncorrelatedness and independence are equivalent. In a local Gaussian
approximation, this property is used to asses dependence by means of the
local correlation, and tests of independence are constructed by accumulating
the local Gaussian correlation, see Chapter 7.
Another very important property of the Gaussian distribution is that
marginal distributions and conditional distributions are again Gaussian. Let
X ∼ N (μX , XX ) be a p-dimensional column vector, and let Y be a q-
dimensional column vector with mean μY = E(Y ) and covariance matrix
Parametric, nonparametric, locally parametric 11
YY . Assume that the covariance matrix XX = E(X − μX )(X − μX )T

is nonsingular, and assume that (X , Y ) is jointly (p + q)-variate normally
distributed. Then the conditional distribution of Y given X is Gaussian
with mean
−1
μY |X =x = μY + YX XX (x − μX ) (2.2)
with YX = E(Y − μY )(X − μX )T , and with covariance matrix
−1
Y |X = YY − YX XX XY . (2.3)
In the bivariate case X = (X1 , X2 ), this reduces to

σ2
μ2|1 = μ2 + ρ21 (x1 − μ1 )
σ1
and
2|1 = (1 − ρ21
2
)σ22 .
These formulas are starting points for defining the partial correlation func-
tion of two vectors X and Y given a third vector Z. In Chapter 11, we will
introduce local versions of these quantities by straight analogy and use them
in a description of conditional density functions and tests for conditional
independence. The local partial correlation function derived from the for-
mulas for the global Gaussian will play an essential role in these derivations.
Other useful properties of a multivariate Gaussian are its simple transfor-
mation rules. If X is a multivariate Gaussian of dimension p, c is a vector of
scalars having dimension q, B is a q × p matrix of scalars, and if X ∼ N (μ, ),
then Y = c + BX ∼ N (c + Bμ, B BT ). In particular, any linear combina-
tion of the components of X is again normally distributed. We will make
use of this in Section 6.4 on nonlinear local portfolio construction.
The estimation of the parameters μ and in (2.1) can be done by
maximizing the log likelihood function, which becomes very simple for
the density function (2.1). The analogue for estimating a local mean and
a local covariance is a local log likelihood function as explained in some
detail in Chapter 4.
2.2.2 The elliptical distribution

The normal distribution plays a very central role in this book, since it forms
the basis for a local Gaussian representation. For a distribution that is not
normal but belongs to a specific class of distributions, say the t-distribution,
it may of course be a disadvantage to try to approximate it with a family

of Gaussian distributions. Then the question arises whether we should try
to use a family of more general distributions. In particular, we might be
interested in distributions that retain as many as possible of the very simple
properties of Gaussian distributions and that are easily extendable to the
multivariate case. A natural class for this purpose is the elliptical distribu-
tions. This class of distributions is also of interest in its own right when
it comes to applications in finance, since it contains several heavy-tailed
distributions. A classic paper in this context is by Owen and Rabinovitch
(1983). A more recent contribution is by Landsman and Valdez (2003).
There is also a separate chapter on elliptical distributions in McNeil et al.
(2005).
There are several ways to define an elliptic distribution, which can be
defined for both discrete and continuous variables. Perhaps the simplest
definition is taking as a starting point the class of spherical distributions. The
random q-dimensional variable X is said to have a spherical distribution if
its characteristic function φX (t) can be written as

φX (t) = E exp(itT X ) = ψ tT t = ψ t12 + · · · + tq2
for some function ψ , which is called the characteristic generator. We write

X ∼ Sq (ψ). A particular case of spherical multivariate distributions is ob-
tained by taking ψ(t) = exp(−tT t/2), which is the multivariate normal
X ∼ N (0, I q ) with I q being the q-dimensional identity matrix.
Then the general p-dimensional distribution is defined by the distribu-
tion of a stochastic variable X of dimension p given by
X = μ + AX ,
where μ is a p-dimensional vector of real numbers, A is a p × q matrix, and

X ∼ Sq (ψ).
Hence we obtain elliptical distributions by multivariate affine transfor-
mations of spherical distributions. The characteristic function is

φX (t) = E exp itT X

= E exp itT μ + AX
= exp(itT μ)E(exp(i(AT t)T )X )
Tμ
= eit ψ(tT t), (2.4)
Another random document with
no related content on Scribd:
'You describe him. He charges like a bull. He bore off the prize
that day against all comers. The Lord of Genestra had his thigh
broken by him.'
'So, so!' said Bellarion, very thoughtful. 'It's my neck he means to
break to-morrow. I read it in his smile.'
'A swaggerer,' said Stoffel. 'He'll take a heavy fall one day.'
'Unfortunately that day is not to-morrow.'
'Are you to ride against him, then?' There was concern in Stoffel's
voice.
'So he believes. But I don't. I have a feeling that to-morrow I shall
not be in case to ride against any one. I have a fever coming on: the
result of hardships suffered on the way from Travo. Nature will
compel me, I suspect, to keep my bed to-morrow.'
Stoffel considered him with grave eyes. 'Are you afraid?'
'What else?'
'And you confess it?'
'It asks courage. Which shows that whilst afraid I am not a
coward. Life is full of paradox, I find.'
Stoffel laughed. 'No need to protest your courage to me. I
remember Travo.'
'There I had a chance to succeed. Here I have none. And who
accepts such odds is not a brave man, but a fool. I don't like broken
bones; and still less a broken reputation. I mean to keep what I've
won against the day when I may need it. Reputation, Stoffel, is a
delicate bubble, easily pricked. To be unhorsed in the lists is no
proper fate for a hero.'
'You're a calculating rogue!'
'That is the difference between me and Carmagnola, who is just a
superior man-at-arms. Each to his trade, Werner, and mine isn't of
the tilt-yard, however many knighthoods they bestow on me. Which
is why to-morrow I shall have the fever.'
This resolve, however, went near to shipwreck that same evening.
In the Hall of Galeazzo the Duke gave audience, which was to be
followed by a banquet. Bidden to this came the new knight Bellarion,
trailing a splendid houppelande of sapphire velvet edged with
miniver that was caught about his waist by a girdle of hammered
silver. He had dressed himself with studied care in the azure and
argent of his new blazon. His tunic, displayed at the breast, where
the houppelande fell carelessly open, and at the arms which
protruded to the elbow from the wide short sleeves of his upper
garment, was of cloth of silver, whilst his hose was in broad vertical
stripes of alternating blue and white. Even his thick black hair was
held in a caul of fine silver thread that was studded with sapphires.
Imposingly tall, his youthful lankness dissembled by his dress, he
drew the eyes of the court as he advanced to pay homage to the
Duke.
Thereafter he was held awhile in friendly talk by della Torre and
the Archbishop. It was in escaping at last from these that he found
himself suddenly looking into the solemn eyes of the Princess
Valeria, of whose presence in Milan this was his first intimation.
She stood a little apart from the main throng under the fretted
minstrel's gallery, at the end of the long hall, with the handsome
Monna Dionara for only companion.
Startled, he turned first red, then white, under the shock of that
unexpected encounter. He had a feeling, under those inscrutable
eyes, of being detected, stripped of his fine trappings and audacious
carriage, and discovered for an upstart impostor, the son of nobody,
impudently ruffling it among the great.
Thus an instant. Then, recovering his poise, he went forward with
leisurely dignity to make his bow, in which there was nothing rustic.
She coloured slightly. Her eyes kindled, and she drew back as if
to depart. A single interjectory word escaped her: 'Audacious!'
'Lady, I thank you for the word. It shall supply the motto I still lack:
"Audax," remembering that "Audaces fortuna juvat."'
She had not been a woman had she not answered him.
'Fortune has favoured you already. You prosper, sir.'
'By God's grace, madonna.'
'God has less to do with it, I think, than your own arts.'
'My arts?' He questioned not the word, but the meaning she
applied to it.
'Such arts as Judas used. You should study the end he made.'
On that she would have gone, but the sharpness of his tone
arrested her.
'Madonna, if ever I practised those arts, it was in your service, and
a reproach is a poor requital.'
'In my service!' Her eyes momentarily blazed. 'Was it in my
service that you came to spy upon me and betray me? Was it in my
service that you murdered Enzo Spigno?' She smiled with terrible
bitterness. 'I have, you see, no illusions left of the service that you
did me.'
'No illusions!' His voice was wistful. She reasoned much as he
had feared that she would reason. 'Lord God! You are filled with
illusions; the result of inference; and I warned you, madonna, that
inference is not your strength.'
'You poor buffoon! Will you pretend that you did not murder
Spigno?'
'Of course I did.'
The admission amazed her where she had expected denial.
'You confess it? You dare to confess it?'
'So that in future you may assert with knowledge what you have
not hesitated to assert upon mere suspicion. Shall I inform you of the
reason at the same time? I killed Count Spigno because he was the
spy sent by your uncle to betray you, so that your brother's ruin
might be accomplished.'
'Spigno!' she cried in so loud a voice of indignation that her lady
clutched her arm to impose caution. 'You say that of Spigno? He was
the truest, bravest friend I ever knew, and his murder shall be atoned
if there is a justice in heaven. It is enough.'
'Not yet, madonna. Consider only that one circumstance which
intrigued the Podestà of Casale: that at dead of night, when all
Barbaresco's household was asleep, only Count Spigno and I were
afoot and fully dressed. Into what tale does that fit besides the lie I
told the Podestà? Shall I tell you?'
'Shall I listen to one who confesses himself a liar and murderer?'
'Alas! Both: in the service of an ungracious lady. But hear now the
truth.'
Briefly and swiftly he told it.
'I am to believe that?' she asked him in sheer scorn. 'I am to be so
false to the memory of one who served me well and faithfully as to
credit this tale of his baseness upon no better word than yours?
Why, it is a tale which even if true must brand you for a beast. This
man, whatever he may have been, was moved to rescue you, you
say, from certain doom; and all the return you made him for that act
of charity was to stab him!'
He wrung his hands in despair. 'Oh, the perversity of your
reasoning! But account me a beast if you will for the deed. Yet admit
that the intention was selfless. Judge the result. I killed Count Spigno
to make you safe, and safe it has made you. If I had other aims, if I
were an agent to destroy you, why did I not speak out in the
Podestà's court?'
'Because your unsupported word would hardly have sufficed to
doom persons of our condition.'
'Which again is precisely why I killed Count Spigno: because if he
had lived, he would have supported it. Is it becoming clear?'
'Clear? Shall I tell you what is clear? That you killed Spigno in
self-defence when he discovered you for the Judas that you were.
Oh, believe me, it is very clear. To make it so there are your lies to
me, your assertion that you were a poor nameless scholar who had
imposed himself upon the Marquis Theodore by the pretence of
being Facino Cane's son. A pretence you said it was. You'll deny that
now.'
Some of his assurance left him. 'No. I don't deny it.'
'You'll tell me, perhaps, that you deceived the Lord Facino himself
with that pretence?' And now without waiting for an answer, she
demolished him with the batteries of her contempt. 'In so great a
pretender even that were possible. You pretended to lay down your
life at Travo, yet behold you resurrected to garner the harvest which
that trick has earned you.'
'Oh, shameful!' he cried out, stirred to anger by a suspicion so
ignoble.
'Are you not rewarded and knighted for the stir that was made by
the rumour of your death? You are to give proof of your knightly
worth in the lists to-morrow. It will be interesting.'
On that she left him standing there with wounds in his soul that
would take long to heal. When at last he swung away, a keen eye
observed the pallor of his face and the loss of assurance from his
carriage; the eye of Facino's lady who approached him on her lord's
arm.
'You are pale, Bellarion,' she commented in pure malice, having
watched his long entertainment with the Princess of Montferrat.
'Indeed, madonna, I am none so well.'
'Not ailing, Bellarion?' There was some concern in Facino's tone
and glance.
And there and then the rogue saw his opportunity and took it.
'It will be nothing.' He passed a hand across his brow.
'The excitement following upon the strain of these last days.'
'You should be abed, boy.'
'It is what I tell myself.'
He allowed Facino to persuade him, and quietly departed. His
sudden illness was rumoured later at the banquet when his place
remained vacant, and consequently there was little surprise when it
was known on the morrow that a fever prevented him from bearing
his part in the jousts at Porta Giovia.
By the doctor who ministered to him, he sent a message to
Carmagnola of deepest and courtliest regret that he was not
permitted to rise and break a lance with him.
CHAPTER XI
THE SIEGE OF ALESSANDRIA
Gabriello Maria Visconti's plans for the restoration of Ghibelline

authority suffered shipwreck, as was to be expected in a council
mainly composed of Guelphs.
The weapon placed in their hands by Gabriello Maria for his own
defeat was the Marquis Theodore's demand, as the price of his
alliance, that he should be supported in the attempt to recover
Genoa to Montferrat.
Della Torre laughed the proposal to scorn. 'And thereby incur the
resentment of the King of France!' He developed that argument so
speciously that not even Facino, who was present, suspected that it
did not contain the true reason of della Torre's opposition.
In hiring a French contingent to strengthen the army which he had
led against Buonterzo, Facino had shown the uses that could be
made of Boucicault. What Facino had done della Torre could do,
nominally on the Duke's behalf. He could hire lances from Boucicault
to set against Facino himself when the need for this arose.
'Possibly,' ventured Gabriello, 'the surrender of Vercelli and certain
other guarantees would suffice to bring Montferrat into alliance.'
But della Torre desired no such alliance. 'Surrender Vercelli! We
have surrendered too much already. It is time we sought alliances
that will restore to Milan some of the fiefs of which she has been
robbed.'
'And where,' Facino quietly asked him, 'will you find such allies?'
Della Torre hesitated. He knew as well as any man that policies
may be wrecked by premature disclosure. If his cherished scheme of
alliance with Malatesta of Rimini were suspected, Facino,
forewarned, would arm himself to frustrate it. He lowered his glance.
'I am not prepared to say where they may be found. But I am
prepared to say that they are not to be found in Theodore of
Montferrat at the price demanded by that Prince.'
Gabriello Maria was left to make what excuses he could to the
Marquis Theodore; and the Marquis Theodore received them in no
pleasant manner. He deemed himself slighted, and said so; hinting
darkly that Milan counted enemies enough already without wantonly
seeking to add to them. Thus in dudgeon he returned to Montferrat.
Della Torre's patient reticence was very shortly justified.
In the early days of June came an urgent and pitiful appeal from
the Duke's brother, Filippo Maria, Count of Pavia, for assistance
against the Vignati of Lodi, who were ravaging his territories and had
seized the city of Alessandria.
The Duke was in his closet with della Torre and Lonate when that
letter reached him. He scowled and frowned and grunted over the
parchment awhile, then tossed it to della Torre.
'A plague on him that wrote it! Can you read the scrawl, Antonio?'
Della Torre took it up. 'It is from your brother, highness; the Lord
Filippo Maria.'
'That skin of lard!' Gian Maria was contemptuous. 'If he
remembers my existence, he must be in need of something.'
Della Torre gravely read the letter aloud. The Prince guffawed
once or twice over a piteous phrase, meanwhile toying with the head
of a great mastiff that lay stretched at his feet.
He guffawed more heartily than ever at the end, the malice of his
nature finding amusement in the calamities of his brother. 'His
Obesity of Pavia is disturbed at last! Let the slothful hog exert
himself, and sweat away some of his monstrous bulk.'
'Do not laugh yet, my lord.' Della Torre's lean, crafty, swarthy face
was grave. 'I have ever warned you against the ambition of Vignate,
and that it would not be satisfied with the reconquest of Lodi. He is in
arms, not so much against your brother as against the house of
Visconti.'
'God's bones!' Goggle-eyed, the Duke stared at his adviser. Then
to vent unreasoning fury he rose and caught the dog a vicious kick
which drove it yelping from him. 'By Hell, am I to go in arms against
Vignate? Is that your counsel?'
'No less.'
'And this campaign against Buonterzo scarcely ended! Am I to
have nothing but wars and feuds and strife to distract my days? Am I
to spend all in quelling brigandage? By the Passion! I'd as soon be
Duke of Hell as reign in Milan.'
'In that case,' said della Torre, 'do nothing, and the rest may
follow.'
'Devil take you, Antonio!' He caught up a hawk-lure from the table,
and set himself to strip it as he talked, scattering the feathers about
the room. 'Curb him, you say? Curb this damned thief of Lodi? How
am I to curb him? The French lances are gone back to Boucicault.
The parsimonious fathers of this miserly city were in haste to dismiss
them. They think of nothing but ducats, may their souls perish! They
think more of ducats than of their duke.' Inconsequently, peevishly,
he ranted on, reducing the hawk-lure to rags the while, and showing
the crafty della Torre his opportunity.
'Vignate,' he said at last, when the Duke ceased, 'can be in no
great strength when all is reckoned. Facino's own condotta should
fully suffice to whip him out of Alessandria and back to Lodi.'
Gian Maria moved restlessly about the room.
'What if it should not? What if Facino should be broken by
Vignate? What then? Vignate will be at the gates of Milan.'
'He might be if we could not prepare for the eventuality.'
With a sudden curious eagerness Gian Maria glared at his
mentor. 'Can we? In God's name, can we? If we could ...' He
checked. But the sudden glow of hate and evil hope in his prominent
pale eyes showed how he was rising to the bait.
Della Torre judged the moment opportune. 'We can,' he answered
firmly.
'How, man? How?'
'In alliance with Malatesta your highness would be strong enough
to defy all comers.'
'Malatesta!' The Duke leapt as if stung. But instantly he curbed
himself. The loose embryonic features tightened, reflecting the
concentration of the embryonic wicked mind within. 'Malatesta, eh?'
His tone was musing. He let himself drop once more into his broad
armchair, and sat there, cross-legged, pondering.
Della Torre moved softly to his side, and lowered his voice to an
impressive note.
'Indeed, your highness should consider whether you will not in
any event bring in Malatesta so soon as Facino has departed on this
errand.'
The handsome, profligate Lonate, lounging, a listener by the
window, cleared up all ambiguity: 'And so make sure that this upstart
does not return to trouble you again.'
Gian Maria's head sank a little between his shoulders. Here was
his chance to rid himself for all time of the tyrannical tutelage of that
condottiero, made strong by popular support.
'You speak as if sure that Malatesta will come.'
Della Torre put his cards on the table at last. 'I am. I have his word
that he will accept a proposal of alliance from your highness.'
'You have his word!' The ever-ready suspicions of a weak mind
were stirring.
'I took his feeling against the hour when your potency might need
a friend.'
'And the price?'
Della Torre spread his hands. 'Malatesta has ambitions for his
daughter. If she were Duchess of Milan ...'
'Is that a condition?' The Duke's voice was sharp.
'A contingency only,' della Torre untruthfully assured him. 'Yet if
realised the alliance would be consolidated. It would become a
family affair.'
'Give me air! Let me think.' He rose, thrusting della Torre away by
a sweep of his thin arm.
Ungainly in his gaudy red and white, shuffling his feet as he went,
he crossed to the window where Lonate made way for him. There he
stood a moment looking out, whilst between Lonate and della Torre a
look of intelligence was flashed.
Suddenly the boy swung round again, and his grotesque
countenance was flushed. 'By God and His Saints! What thought
does it ask?' He laughed, slobberingly, at the picture in his mind of a
Facino Cane ruined beyond redemption. Nor could he perceive, poor
fool, that he would be but exchanging one yoke for another, probably
heavier.
Still laughing, he dismissed della Torre and Lonate, and sent for
Facino. When the condottiero came, he was given Filippo Maria's
letter, which he spelled out with difficulty, being little more of a
scholar than the Duke.
'It is grave,' he said when he had reached the end.
'You mean that Vignate is to be feared?'
'Not so long as he is alone. But how long will he so continue?
What if he should be joined by Estorre Visconti and the other
malcontents? Singly they matter nothing. United they become
formidable. And this bold hostility of Vignate's may be the signal for a
league.'
'What then?'
'Smash Vignate and drive him out of Alessandria before it
becomes a rallying-ground for your enemies.'
'About it, then,' rasped the Duke. 'You have the means.'
'With the Burgundians enlisted after Travo, my condotta stands at
two thousand three hundred men. If the civic militia is added ...'
'It is required for the city's defence against Estorre and the other
roving insurgents.'
Facino did not argue the matter.
'I'll do without it, then.'
He set out next day at early morning, and by nightfall, the half of
that march to Alessandria accomplished, he brought his army,
wearied and exhausted by the June heat, to rest under the red walls
of Pavia.
To proceed straight against the very place which Vignate had
seized and held was a direct course of action in conflict with ideas
which Bellarion did not hesitate to lay before the war-experienced
officers composing Facino's council. He prefaced their exposition by
laying down the principle, a little didactically, that the surest way to
defeat an opponent is to assault him at the weakest point. So much
Facino and his officers would have conceded on the battle-ground
itself. But Bellarion's principle involved a wider range, including the
enemy's position before ever battle was joined so as to ensure that
the battle-ground itself should be the enemy's weakest point. The
course he now urged entailed an adoption of the strategy employed
by the Athenians against the Thebans in the Peloponnesian war, a
strategy which Bellarion so much admired and was so often to apply.
In its application now, instead of attacking Alessandria behind
whose walls the enemy lay in strength, he would have invaded
Vignate's own temporarily unguarded Tyranny of Lodi.
Facino laughed a little at his self-sufficiency, and, emboldened by
that, Carmagnola took it upon himself to put the fledgling down.
'It is in your nature, I think, to avoid the direct attack.' He sneered
as he spoke, having in mind the jousts at Milan and the manner in
which Bellarion had cheated him of the satisfaction upon which he
counted. 'You forget, sir, that your knighthood places you under
certain obligations.'
'But not, I hope,' said Bellarion innocently, 'under the obligation of
being a fool.'
'Do you call me that?' Carmagnola's sudden suavity was in itself a
provocation.
'You boast yourself the champion of the direct attack. It is the
method of the bull. But I have never heard it argued from this that the
bull is intelligent even among animals.'
'So that now you compare me with a bull?' Carmagnola flushed a
little, conscious that Koenigshofen and Stoffel were smiling.
'Quiet!' growled Facino. 'We are not here to squabble among
ourselves. Your assumptions, Bellarion, sometimes become
presumptions.'
'So you thought on the Trebbia.'
Facino brought his great fist down upon the table. 'In God's name!
Will you be pert? You interrupt me. Battering-ram tactics are not in
my mind. I choose a different method. But I attack Alessandria none
the less, because Vignate and his men are there.'
Discreetly Bellarion said no more, suppressing the argument that
by reducing unguarded Lodi and restoring it to the crown of Milan
from which it had been ravished, a moral effect might be produced of
far-reaching effect upon the fortunes of the duchy.
After a conference with Filippo Maria in his great castle of Pavia,
Facino resumed his march, his army now increased by six hundred
Italian mercenaries under a soldier of fortune named Giasone Trotta,
whom Filippo Maria had hired. He took with him a considerable train
of siege artillery, of mangonels, rimbaults, and cannon, to which the
Count of Pavia had materially added.
Nevertheless, he did not approach Alessandria within striking
distance of such weapons. He knew the strength to withstand
assault of that fortress-city, built some three hundred years before on
the confines of the Pavese and Montferrat to be a Guelphic
stronghold in the struggle between Church and Empire. Derisively
then the Ghibellines had dubbed it a fortress of straw. But astride of
the river Tanaro, above its junction with the Bormida, this
Alessandria of Straw had successfully defied them.
Facino proposed to employ the very strength of her strategic
position for the undoing of her present garrison if it showed fight. And
meanwhile he would hem the place about, so as to reduce it by
starvation.
Crossing the Po somewhere in the neighbourhood of Bassignana,
he marched up the left bank of the Tanaro to Pavone, a village in the
plain by the river just within three miles of Alessandria. There he took
up his quarters, and thence on a radius of some three miles he drew
a cordon throughout that low-lying, insalubrious land, intersected
with watercourses, where only rice-fields flourished. This cordon
crossed the two rivers just above their junction, swept thence to
Marengo, recrossing the Bormida, ran to Aulara in the south and on
to Casalbagliano in the West, just beyond which it crossed the
Tanaro again, and, by way of San Michele in the north, went on to
complete the circle at Pavone.
So swift had been the movement that the first intimation to the
Alessandrians that they were besieged was from those who, issuing
from the city on the morrow, were stopped at the lines and ordered to
return.
From information obtained from these, in many cases under threat
of torture, it became clear that the populous city was indifferently
victualled, and unequal, therefore, to a protracted resistance. And
this was confirmed during the first week by the desperate efforts
made by Vignate, who was raging like a trapped wolf in Alessandria.
Four times he attempted to break out in force. But within the outer
circle, and close to the city so as to keep it under observation,
Facino had drawn a ring of scouts, whose warning in each case
enabled him to concentrate promptly at the point assailed. The
advantage lay with Facino in these engagements, since the cavalry
upon which Vignate chiefly depended found it impossible to operate
successfully in those swampy plains. Over ground into which the
horses sank to their fetlocks at every stride, a cavalry charge was a
brutum fulmen. Horses were piked by Koenigshofen's foot, and
formations smashed and hurled back by an enemy upon whom their
impact was no more than a spent blow.
If they escaped it was because Facino would make no prisoners.
He would not willingly relieve Alessandria of a single mouth that
would help to eat up its power of endurance. For the same reason he
enjoined it upon his officers that they should be as sparing as
possible of life.
'That is to say, of human life,' said Bellarion, raising his voice in
council for the first time since last rebuked.
They looked at him, not understanding.
'What other life is in question?' asked Carmagnola.
'There are the horses. If allowed to survive, they may be eaten in
the last extremity.'
They acted upon that reminder when Vignate made his next sally.
Facino did not wait as hitherto to receive the charge upon his pikes,
but raked the enemy ranks, during their leisurely advance and again
during their subsequent retreat with low-aimed arbalest bolts which
slew only horses.
Whether Vignate perceived the reason, or whether he came to
realise that the ground was not suitable for cavalry, his fourth sally, to
the north in the direction of San Michele, was made on foot. He had
some two thousand men in his following, and had they been lightly
armed and properly led it is probable that they would have broken
through, for the opposing force was materially less. But Vignate,
unaccustomed to handling infantry, committed the error of the
French at Agincourt. He employed dismounted men-at-arms in all
the panoply in which normally they rode to battle. Their fate was
similar to that of the French on that earlier occasion. Toiling over the
clammy ground in their heavy armour, their advance became leaden-
footed, and by the time they reached Facino's lines they were
exhausted men easily repulsed, and as glad as they were surprised
to escape death or capture.
After that failure, three representatives of the Commune of
Alessandria, accompanied by one of Vignate's captains, presented
themselves at Facino's quarters in the house of the Curate of
Pavone, temporarily appropriated by the condottiero.
They were ushered into a plain yellow-washed room, bare of all
decoration save that of a crudely painted wooden crucifix which hung
upon the wall above a straight-backed wooden settle. An oblong
table of common pine stood before this settle; a writing-pulpit, also of
pine, placed under one of the two windows by which the place was
lighted, and four rough stools and a shallow armchair completed the
furniture.
The only gentle touch about that harsh interior was supplied by
the sweet-smelling lemon verbena and rosemary mingled in the
fresh rushes with which the floor was copiously strewn to dissemble
its earthen nudity.
Carmagnola, showily dressed as usual in blue and crimson, with
marvellously variegated hose and a jewelled caul confining his flaxen
hair, had appropriated the armchair, and his gorgeous presence
seemed to fill the place. Stoffel, Koenigshofen, Giasone Trotta, and
Vougeois, who commanded the Burgundians, occupied the stools
and afforded him a sober background. Bellarion leaned upon the
edge of the settle, where Facino sat alone, square-faced and stern,
whilst the envoys invited him to offer terms for the surrender of the
city.
'The Lord Count of Pavia,' he told them, 'does not desire to mulct
too heavily those of his Alessandrian subjects who have remained
loyal. He realises the constraint of which they may have been the
victims, and he will rest content with a payment of fifty thousand
florins to indemnify him for the expenses of this expedition.' The
envoys breathed more freely. But Facino had not yet done. 'For
myself I shall require another fifty thousand florins for distribution
among my followers, to ransom the city from pillage.'
The envoys were aghast. 'One hundred thousand gold florins!'
cried one. 'My lord, it will ...'
He raised his hand for silence. 'That as regards the Commune of
Alessandria. Now, as concerns the Lord Vignate, who has so rashly
ventured upon this aggression. He is allowed until noon to-morrow to
march out of Alessandria with his entire following, but leaving behind
all arms, armour, horses, bullocks, and war material of whatsoever
kind. Further, he will enter into a bond for one hundred thousand
florins, to be paid either by himself personally or by the Commune of
Lodi to the Lord Count of Pavia's city of Alessandria, to indemnify the
latter for the damages sustained by this occupation. And my Lord
Vignate will further submit to the occupation of the city of Lodi by an
army of not more than two thousand men, who will be housed and
fed and salaried at the city of Lodi's charges until the indemnity is
paid. With the further condition that if payment is not made within
one month, the occupying army shall take it by putting the city to
sack.'
The officer sent by Vignate, a stiff, black-bearded fellow named
Corsana, flushed indignantly. 'These terms are very harsh,' he
complained.
'Salutary, my friend,' Facino corrected him. 'They are intended to
show the Lord Vignate that brigandage is not always ultimately
profitable.'
'You think he will agree?' The man's air was truculent. The three
councillors looked scared.
Facino smiled grimly. 'If he has an alternative, let him take
advantage of it. But let him understand that the offer of these terms
is for twenty-four hours only. After that I shall not let him off so
lightly.'
'Lightly!' cried Corsano in anger, and would have added more but
that Facino cropped the intention.
'You have leave to go.' Thus, royally, Facino dismissed them.
They did not return within the twenty-four hours, nor as day
followed day did Vignate make any further sign. Time began to hang
heavily on the hands of the besiegers, and Facino's irritation grew
daily, particularly when an attack of the gout came to imprison him in
the cheerless house of the Curate of Pavone.
One evening a fortnight after the parley and nearly a month after
the commencement of the siege, as Facino sat at supper with his
officers, all save Stoffel, who was posted at Casalbagliano, the
condottiero, who was growing impatient of small things, inveighed
against the quality of the food.
It was Giasone Trotta, to whose riders fell the task of provisioning
the army, who answered him. 'Faith! If the siege endures much
longer, it is we who will be starved by it. My men have almost
cleaned up the countryside for a good ten miles in every direction.'
It was a jocular exaggeration, but it provoked an explosion from
Facino.
'God confound me if I understand how they hold out. With two
thousand ravenous soldiers in the place, a week should have
brought them to starvation.'
Koenigshofen thoughtfully stroked his square red beard. 'It's
colossally mysterious,' said he.
'Mysterious, aye! That's what plagues me. They must be fed from
outside.'
'That is quite impossible!' Carmagnola was emphatic. As Facino's
lieutenant, it fell to his duty to see that the cordon was properly
maintained.
'Yet what is the alternative,' wondered Bellarion, 'unless they are
eating one another?'
Carmagnola's blue eyes flashed upon him almost malevolently for
this further reflection upon his vigilance.
'You set me riddles,' he said disdainfully.
'And you're not good at riddles, Francesco,' drawled Bellarion,
meeting malice with malice. 'I should have remembered it.'
Carmagnola heaved himself up. 'Now, by the Bones of God, what
do you mean?'
The ears of the ill-humoured Facino had caught a distant sound.
'Quiet, you bellowing calf!' he snapped. 'Listen! Listen! Who comes
at that breakneck speed?'
It was a hot, breathless night of July, and the windows stood wide
to invite a cooling draught. As the four men, so bidden, grew
attentive, they caught from the distance the beat of galloping hooves.
'It's not from Alessandria,' said Koenigshofen.
'No, no,' grunted Facino, and thereafter they listened in silence.
There was no reason for it save such colour as men's imaginings
will give a sound breaking the deathly stillness of a hot dark night,
yet each conceived and perhaps intercommunicated a feeling that
these hooves approaching so rapidly were harbingers of portents.
Carmagnola went to the door as two riders clattered down the
village street, and, seeing the tall figure silhouetted against the light
from within, they slackened pace.
'The Lord Facino Cane of Biandrate? Where is he quartered?'
'Here!' roared Carmagnola, and at the single word the horses
were pulled up with a rasping of hooves that struck fire from the
ground.
CHAPTER XII
VISCONTI FAITH
If Facino Cane's eyes grew wide in astonishment to see his

countess ushered into that mean chamber by Carmagnola, wider still
did they grow to behold the man who accompanied her and to
consider their inexplicable conjunction. For this man was Giovanni
Pusterla of Venegono, cousin to that Pusterla who had been
castellan of Monza, and who by Gian Maria's orders had procured
the assassination of Gian Maria's mother.
The rest is a matter of history upon which I have already touched.
In a vain attempt to mask his own matricide, to make the crime
appear as the work of another, Gian Maria had seized the
unfortunate castellan who had served his evil will too faithfully and
charging him with the crime caused him barbarously and without trial
to be done to death. Thereafter, because he perceived that this did
not suffice to turn the public mind from the conviction of his own
horrible guilt, Gian Maria had vowed the extermination of the
Pusterla family, as a blood-offering to the manes of his murdered
mother. It was a Pusterla whom he had hunted with his dogs into the
arms of Bellarion in the meadows of Abbiategrasso, and that was the
fifth innocent member of the family whom he had done to death in
satisfaction of his abominable vow.
This Pusterla of Venegono, who now led the Countess Beatrice
into her husband's presence, was a slight but vigorous and
moderately tall man of not more than thirty, despite the grey that so
abundantly mingled with his thick black hair. His shaven
countenance was proud and resolute, with a high-bridged nose
flanked perhaps too closely by dark eyes that glowed and flashed as
in reflection of his superabundant energy of body and of spirit.
Between himself and Facino there was esteem; but no other link
to account for his sudden appearance as an escort to the Lady
Beatrice.
From the settle which he occupied, his ailing leg stretched upon it,
the amazed Facino greeted them by a rough soldier's oath on a note
of interrogation.
The Countess, white and lovely, swept towards him.
'You are ailing, Facino!' Concern charged her murmuring voice as
she stooped to receive his kiss.
His countenace brightened, but his tone was almost testy.
To discuss his ailments now was but to delay the explanation that
he craved. 'That I ail is no matter. That you should be here ... What
brings you, Bice, and with Venegono there?'
'Aye, we take you by surprise,' she answered him. 'Yet Heaven
knows there would be no need for that if ever you had heeded me, if
ever you had used your eyes and your wits as I bade you.'
'Will you tell me what brings you, and leave the rest?'
She hesitated a moment, then swung imperially to her travelling
companion.
'Tell him, Messer da Venegono.'
Venegono responded instantly. He spoke rapidly, using gestures
freely, his face an ever-shifting mirror of his feelings, so that at once
you knew him for a brisk-minded, impulsive man. 'We are here to
speak of what is happening in Milan. Do you know nothing of it, my
lord?'
'In Milan? Despatches reach me weekly from his highness. They
report nothing that is not reassuring.'
The Countess laughed softly, bitterly. Venegono plunged on.
'Is it reassuring to you that the Malatesta of Rimini, Pandolfo, and
his brother Carlo are there with an army five thousand strong?'
Facino was genuinely startled. 'They are moving against Milan?'
Again the Countess laughed, and this time Venegono laughed
with her.

(Download PDF) Statistical Modelling Using Local Gaussian Approximation 1St Edition Tjostheim Ebook Online Full Chapter

Uploaded by

Copyright:

Available Formats

You might also like

(Download PDF) Statistical Modelling Using Local Gaussian Approximation 1St Edition Tjostheim Ebook Online Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Download PDF) Statistical Modelling Using Local Gaussian Approximation 1St Edition Tjostheim Ebook Online Full Chapter

Uploaded by

Copyright:

Available Formats

Statistical Modelling using Local

Gaussian Approximation 1st Edition

From Statistical Physics to Data-Driven Modelling

Heat Transfer Modelling Using COMSOL 1st Edition Layla

Statistical Data Analysis Using SAS Intermediate

Asymptotic Statistical Inference A Basic Course Using R

Gaussian Measures in Finite and Infinite Dimensions 1st

Computation and Approximation 1st Edition Vijay Gupta

Marketing Analytics Statistical Tools for Marketing and

The Psychosocial Reality of Digital Travel Being in

Library of Congress Cataloging-in-Publication Data

British Library Cataloguing-in-Publication Data

For information on all Academic Press publications

Publisher: Candice Janco

2. Parametric, nonparametric, locally parametric 7

4. Local Gaussian correlation and dependence 87

4.10. Proof of asymptotic results 123

5. Local Gaussian correlation and the copula 135

6. Applications in ﬁnance 161

7. Measuring dependence and testing for independence 213

8. Time series dependence and spectral analysis 261

9. Multivariate density estimation 301

9.8. Proofs 327

10. Conditional density estimation 335

11. The local Gaussian partial correlation 353

12. Regression and conditional regression quantiles 385

13. A local Gaussian Fisher discriminant 403

Author index 429

The central idea of this book is the approximation of a general multivariate

Chapters 9 and 10 are devoted to estimation of multivariate density

The most important distribution in statistics is the Gaussian distribu-

as is inherent in the Gaussian distribution. In fact, assuming a linear Gaus-

density f . For instance, for a thick-tailed distribution, it can be locally ap-

discuss ways of bypassing it. Somewhat similarly to the additive approxi-

1.1 Computer code

1 See https://github.com/LAJordanger/localgaussSpec for details.

calization of parameters for a quite general parametric family. On the other

2.2 Parametric density models

2.2.1 The Gaussian distribution

We are seeking to replace a given distribution, univariate or multivariate,

where μ = (μ1 , . . . , μp )T with E(Xi ) = μi , and is the covariance matrix

YY . Assume that the covariance matrix XX = E(X − μX )(X − μX )T

In the bivariate case X = (X1 , X2 ), this reduces to

2.2.2 The elliptical distribution

it may of course be a disadvantage to try to approximate it with a family

for some function ψ , which is called the characteristic generator. We write

where μ is a p-dimensional vector of real numbers, A is a p × q matrix, and

THE SIEGE OF ALESSANDRIA

Gabriello Maria Visconti's plans for the restoration of Ghibelline

If Facino Cane's eyes grew wide in astonishment to see his

You might also like