An Introduction to Algebraic Statistics

with Tensors Cristiano Bocci

Visit to download the full and correct content document:
https://textbookfull.com/product/an-introduction-to-algebraic-statistics-with-tensors-cri
stiano-bocci/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

An Introduction to Secondary Data Analysis with IBM

SPSS Statistics First Edition Macinnes

https://textbookfull.com/product/an-introduction-to-secondary-
data-analysis-with-ibm-spss-statistics-first-edition-macinnes/

From Algebraic Structures to Tensors Digital Signal and

Image Processing Matrices and Tensors in Signal
Processing Set 1st Edition Gérard Favier (Editor)

https://textbookfull.com/product/from-algebraic-structures-to-
tensors-digital-signal-and-image-processing-matrices-and-tensors-
in-signal-processing-set-1st-edition-gerard-favier-editor/

An introduction to nonparametric statistics First

Edition Kolassa

https://textbookfull.com/product/an-introduction-to-
nonparametric-statistics-first-edition-kolassa/

An Introduction to Secondary Data Analysis with IBM

SPSS Statistics 1st Edition John Macinnes

https://textbookfull.com/product/an-introduction-to-secondary-
data-analysis-with-ibm-spss-statistics-1st-edition-john-macinnes/
An Algebraic Approach to Symmetry with Applications to
Knot Theory David Edward Joyce

https://textbookfull.com/product/an-algebraic-approach-to-
symmetry-with-applications-to-knot-theory-david-edward-joyce/

MMA320 Introduction to Algebraic Geometry Jan Stevens

https://textbookfull.com/product/mma320-introduction-to-
algebraic-geometry-jan-stevens/

Reasoning with Data An Introduction to Traditional and

Bayesian Statistics Using R 1st Edition Jeffrey M.
Stanton

https://textbookfull.com/product/reasoning-with-data-an-
introduction-to-traditional-and-bayesian-statistics-using-r-1st-
edition-jeffrey-m-stanton/

Statistics for Engineers: An Introduction with Examples

from Practice 1st Edition Hartmut Schiefer

https://textbookfull.com/product/statistics-for-engineers-an-
introduction-with-examples-from-practice-1st-edition-hartmut-
schiefer/

Mathematics in Action: An Introduction to Algebraic,

Graphical, and Numerical Problem Solving (6th Edition)
Consortium For Foundation Mathematics

https://textbookfull.com/product/mathematics-in-action-an-
introduction-to-algebraic-graphical-and-numerical-problem-
solving-6th-edition-consortium-for-foundation-mathematics/
UNITEXT 118 35 28 21 42 7

70 56 42 84 14

25 20 15 30 5

50 40 30 60 10

10 8 6 12 2

20 16 12 24 4

Cristiano Bocci · Luca Chiantini

An Introduction
to Algebraic
Statistics
with Tensors
UNITEXT - La Matematica per il 3+2

Volume 118

Editor-in-Chief
Alfio Quarteroni, Politecnico di Milano, Milan, Italy; EPFL, Lausanne, Switzerland

Series Editors
Luigi Ambrosio, Scuola Normale Superiore, Pisa, Italy
Paolo Biscari, Politecnico di Milano, Milan, Italy
Ciro Ciliberto, Università di Roma “Tor Vergata”, Rome, Italy
Camillo De Lellis, Institute for Advanced Study, Princeton, NJ, USA
Victor Panaretos, Institute of Mathematics, EPFL, Lausanne, Switzerland
Wolfgang J. Runggaldier, Università di Padova, Padova, Italy
The UNITEXT – La Matematica per il 3+2 series is designed for undergraduate
and graduate academic courses, and also includes advanced textbooks at a research
level. Originally released in Italian, the series now publishes textbooks in English
addressed to students in mathematics worldwide. Some of the most successful
books in the series have evolved through several editions, adapting to the evolution
of teaching curricula.

More information about this subseries at http://www.springer.com/series/5418

Cristiano Bocci Luca Chiantini
•

An Introduction to Algebraic
Statistics with Tensors

123
Cristiano Bocci Luca Chiantini
Dipartimento di Ingegneria Dipartimento di Ingegneria
dell’Informazione e Scienze Matematiche dell’Informazione e Scienze Matematiche
Università di Siena Università di Siena
Siena, Italy Siena, Italy

ISSN 2038-5714 ISSN 2532-3318 (electronic)

UNITEXT - La Matematica per il 3+2
ISSN 2038-5722 ISSN 2038-5757 (electronic)
ISBN 978-3-030-24623-5 ISBN 978-3-030-24624-2 (eBook)
https://doi.org/10.1007/978-3-030-24624-2

© Springer Nature Switzerland AG 2019

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, expressed 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.

Cover illustration (LaTeX): A decomposable 3-dimensional tensor of type $3\times 5\times 2$.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
We, the authors, dedicate this book to our
great friend Tony Geramita. When the project
started, Tony was one of the promoters and
he should be among us in the list of authors
of the text. Tony passed away when the book
was at an early stage. We finished the book
following the pattern traced in collaboration
with him, and we always felt as if his
encouragement to continue the project never
faded.
Preface

Statistics and Algebraic Statistics

At the beginning of a book on Algebraic Statistics, it is undoubtedly a good idea to

give the reader some idea of the goals of the discipline.
A reader who is already familiar with the basics of Statistics and Probability is
probably curious about what the prefix “Algebraic” might mean. As we will see,
Algebraic Statistics has its own way of approaching statistical problems, exploiting
algebraic, geometric, or combinatorial properties. These problems are somewhat
different from the one studied by Classical Statistics.
We will illustrate this point of view with some examples, which consider
well-known statistical models and problems. At the same time, we will point out the
difference between the two approaches to these examples.

The Treatment of Random Variables

The initial concern of Classical Statistics is the behavior of one random variable X.
Usually X is identified with a function with values in the real numbers. This is
clearly an approximation. For example, if one records the height of the members of
a population, it is unlikely that the measure goes much further than the second
decimal digit (assume that the unit is 1 m). So, the corresponding graph is a
histogram, with a basic interval of 0:01 m. This is translated to a continuous
variable, by sending the length of the basic interval to zero (and the size of the
population increases).

vii
viii Preface

For random variables of this type, the first natural distribution that one expects is
the celebrated Gaussian distribution, which corresponds to the function

1 ðtlÞ2
XðtÞ ¼ pffiffiffiffiffiffi e 2r2
r 1…

where l and r are parameters which describe the shape of the curve (of course,
other types of distributions are possible, in connection with special behaviors of the
random variable XðtÞ).
The first goal of Classical Statistics is the study of the shape of the function XðtÞ,
together with the determination of its numerical parameters.
When two or more variables are considered in the framework of Classical
Statistics, their interplay can be studied with several techniques. For instance, if we
consider both the heights and the weights of the members of a population and our
goal is a proof of the (obvious) fact that the two variables are deeply connected,
then we can consider the distribution over pairs (height, weight), which is repre-
sented by a bivariate Gaussian, in order to detect the existence of the connection.
The starting point of Algebraic Statistics is quite different. Instead of considering
variables as continuous functions, Algebraic Statistics prefers to deal with a finite
(and possibly small) range of values for the variable X. So, Algebraic Statistics
emphasizes the discrete nature of the starting histogram, and tends to group together
values in wider ranges, instead of splitting them. A distribution over the variable X
is thus identified with a discrete function (to begin with, over the integers).
Algebraic Statistics is rarely interested in situations, where just one random
variable is concerned.
Instead, networks containing several random variables are considered and some
relevant questions raised in this perspective are
• Are there connections between the two or more random variables of the
network?
• Which kind of connection is suggested by a set of data?
• Can one measure the complexity of the connections in a given network of
interacting variables?
Since, from the new point of view, we are interested in determining the relations
between discrete variables, in Algebraic Statistics a distribution over a set of
variables is usually represented by matrices, when two variables are involved, or
multidimensional matrices (i.e., tensors), as the number of variables increases.
It is a natural consequence of the previous discussion that while the main
mathematical tools for Classical Statistics are based on multivariate analysis and
measure theory, the underlying mathematical machinery for Algebraic Statistics is
principally based on the Linear and Multi-linear Algebra of tensors (over the
integers, at the start, but quickly one considers both real and complex tensors).
Preface ix

Relations Among Variables

Just to give an example, let us consider the behavior of a population after the
introduction of a new medicine.
Assume that a population is affected by a disease, which dangerously alters the
value of a glycemic indicator in the blood. This dangerous condition is partially
treated with the new drug. Assume that the purpose of the experiment is to detect
the existence of a substantial improvement in the health of the patients.
In Classical Statistics, one considers the distribution of the random variable X1 ¼
the value of the glycemic indicator over a selected population of patients before the
delivery of the drug, and the random variable X2 ¼ the value of the glycemic
indicator of patients after the delivery of the drug. Both distributions are likely to be
represented by Gaussians, the first one centered at an abnormally high value
of the glycemic indicator, the second one centered at a (hopefully) lower value. The
comparison between the two distributions aims to detect if (and how far) the descent
of the recorded values of the glycemic indicator is statistically meaningful, i.e., if it
can be distinguished from the natural underlying ground noise. The celebrated
Student’s t-test is the world-accepted tool for comparing the means of two Gaussian
distributions and for determining the existence of a statistically significant response.
In many experiments, the response variable is binary or categorical with k levels,
leading to a 2  2, or a 2  k, contingency table. Moreover, when there is more
than one response variable and/or other control variables, the resulting data are
summarized in a multiway contingency table, i.e., a tensor.
This structure may also come from the discretization of a continuous variable.
As an example, consider a population divided into two subsets, one of which is
treated with the drug while the other is treated with traditional methods. Then, the
values of the glycemic indicator are divided into classes (in the roughest case just
two classes, i.e., a threshold which separates two classes is established). After some
passage of time, one records the distribution of the population in the four resulting
categories (treated + under-threshold, treated + over-threshold . . .) which deter-
mines a 2  2 matrix, whose properties encode the existence of a relation between
the new treatment and an improved normalization of the value of the glycemic
indicator (this is just to give an example: in the real world, a much more sophis-
ticated analysis is recommended!).

Bernoulli Binary Models

Another celebrated model, which is different from the Gaussian distribution and is
often introduced at the beginning of a course in Statistics, is the so-called Bernoulli
model over one binary variable.
Assume we are given an object that can assume only two states. A coin, with the
two traditional states H (heads) and T (tails), is a good representation. One has to
x Preface

bear in mind, however, that in the real world, binary objects usually correspond to
biased coins, i.e., coins for which the expected distribution over the two states is not
even.
If p is the probability of obtaining a result (say H) by throwing the coin, then one
can roughly estimate p by throwing the coin several times and determining the ratio

number of throws giving H

total number of throws

but this is usually considered too naïve. Instead, one divides the total set of throws
into several packages, each consisting of r throws, and determines for how many
packages, denoted qðtÞ, one obtained H exactly t times. The value of the constant p
is thus determined by Bernoulli’s formula:
 
r t
qðtÞ ¼ p ð1  pÞrt :
t

By increasing the number of total throws (and thus increasing the number of
packages and the number of throws r in each package), the function qðtÞ tends to a
real function, which can be treated with the usual analytic methods.
Notice that in this way, at the end of the process, the discrete variable Coin is
substituted by a continuous variable qðtÞ. Usually one even goes one step further,
by substituting the variable q with its logarithm, ending up with a linear description.
Algebraic Statistics is scarcely interested in knowing how a single given coin is
biased. Instead, the main goal of Algebraic Statistics is to understand the connec-
tions between the behavior of two coins. Or, better, the connections between the
behavior of a collection of coins.
Consequently, in Algebraic Statistics one defines a collection of variables, one
for each coin, and defines a distribution by counting the records in which the
variables X1 ; X2 ; . . .; Xn have a fixed combination of states. The distribution is
transformed into a tensor of type 2  2      2. All coins can be biased, with
different loads: this does not matter too much. In fact, the main questions that one
expects to solve are
• Are there connections between the outputs of two or more coins?
• Which kind of connection is suggested by the distribution?
• Can one divide the collection of coins into clusters, such that the behavior of
coins of the same cluster is similar?
Answers are expected from an analysis of the associated tensor, i.e., in the
framework of Multi-linear Algebra.
The importance of the last question can be better understood if one replaces
coins with positions in a composite digital signal. Each position has, again, two
possible states, 0 and 1. If the signal is the result of the superposition of many
Preface xi

elementary signals, coming from different sources, and digits coming from the same
source behave similarly, then the division of the signal into clusters yields the
reconstruction of the original message that each source issued.

Splitting into Types

Of course, the separation of several phenomena that are mixed together in a given
distribution is also possible using methods of Classical Statistics.
In a famous analysis of 1894, the biologist Karl Pearson made a statistical study
of the shape of a population of crabs (see [1]). He constructed the histogram for the
ratio between the “forehead” breadth and the body length for 1000 crabs, sampled
in Naples, Italy by W. F. R. Weldon. The resulting approximating curve was quite
different from a Gaussian and presented a clear asymmetry around the average
value. The shape of the function suggested the existence of two distinct types of
crab, each determining its own Gaussian, that were mixed together in the observed
histogram. Pearson succeeded in separating the two Gaussians with the method of
moments. Roughly speaking, he introduced new statistical variables, induced by the
same collection of data, and separated the types by studying the interactions
between the Gaussians of these new variables.
This is the first instance of a computation which takes care of several parameters
of the population under analysis, though the variables are derived from the same set
of data. Understanding the interplay between the variables provides the funda-
mental step for a qualitative description of the population of crabs.
From the point of view of Algebraic Statistics, one could obtain the same
description of the two types which compose the population by adding variables
representing other ratios between lengths in the body of crabs, and analyzing the
resulting tensor.

Mixture Models

Summarizing, Algebraic Statistics becomes useful when the existence and the
nature of the relations between several random variables are explored.
We stress that knowing the shape of the interaction between random variables is
a central problem for the description of phenomena in Biology, Chemistry, Social
Sciences, etc. Models for the description of the interactions are often referred to as
Mixture Models. Thus, mixture models are a fundamental object of study in
Algebraic Statistics.
Perhaps, the most famous and easily described mixture models are the Markov
chains, in which the set of variables is organized in a totally ordered chain, and the
behavior of the variable Xi is only influenced by the behavior of the variable Xi1
(usually, this interaction depends on a given matrix).
xii Preface

Of course, much more complicated types of networks are expected when the
complexity of the collection of variables under analysis increases. So, when one
studies composite signals in the real world, or pieces of a DNA chain, or regions in
a neural tissue, higher level models are likely to be necessary for an accurate
description of the phenomenon.
One thus moves from the study of Markov chains

M1 M2 M3
X1 X2 X3 X4

to the study of Markov trees

X1
M1 M2

X2 X3
M3 M4 M5 M6

X4 X5 X6 X7

and the study of neural nets

M1
X1 X2

M3 M2

X3

In Classical Statistics, the structure of the connections among variables is often a

postulate. In Algebraic Statistics, determining the combinatorics and the topology
of the network is a fundamental task. On the other hand, the time-dependent
activating functions that transfer information from one variable to the next ones,
deeply studied by Classical Statistics, are of no immediate interest for Algebraic
Statistics which, at first, considers steady states of the configuration of variables.
The Multi-linear Algebra behind the aforementioned models is not completely
understood. It requires a deep analysis of subsets of linear spaces described by
parametric or implicit polynomial equations. This is the reason why, at a certain
point, methods of Algebraic Geometry are invoked to push the analysis further.
Preface xiii

Conclusion

The way we think about Algebraic Statistics focuses on aspects of the theory of
random variables which are different from the targets of Classical Statistics. This is
reflected in the point of view introduced in the book. Our general setting differs
from the classical one and is closer to the one implicitly introduced in the books of
Pachter and Sturmfels [2] and Sullivant [3]. Our aim is not to create a new for-
mulation of the whole statistical theory, but only to present an algebraic natural way
in which Statistics can handle problems related to mixture models.
The discipline is currently living in a rapidly expanding network of new insights
and new areas of application. Our knowledge of what we can do in this area is
constantly increasing and it is reasonable to hope that many of the problems
introduced in this book will soon be solved or, if they cannot be solved completely,
then they will at least be better understood. We feel that the time is right to provide
a systematic foundation, with special attention to the application of tensor theory,
for a field that promises to act as a stimulus for mathematical research in Statistics,
and also as a source of suggestions for further developments in Multi-linear Algebra
and Algebraic Geometry.

Siena, Italy Cristiano Bocci

May 2019 Luca Chiantini

Acknowledgements The authors want to warmly thank Fabio Rapallo, who made several fruitful
remarks and suggestions to improve the exposition, especially regarding the connections with
Classical Statistics.

References

1. Pearson K.: Contributions to the mathematical theory of evolution. Phil. Trans. Roy. Soc.
London A, 185, 71–110 (1894)
2. Pachter, L., Sturmfels, B.: Algebraic Statistics for Computational Biology. Cambridge
University Press, New York (2005)
3. Sullivant, S.: Algebraic Statistics. Graduate Studies in Mathematics, vol. 194, AMS,
Providence (2018)
Contents

Part I Algebraic Statistics

1 Systems of Random Variables and Distributions . . . . . . . . . . . . . . . 3
1.1 Systems of Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Measurements on a Distribution . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2 Basic Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Basic Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Booleanization and Logic Connectors . . . . . . . . . . . . . . . . . . . 19
2.3 Independence Connections and Marginalization . . . . . . . . . . . . 21
2.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3 Statistical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Independence Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Connections and Parametric Models . . . . . . . . . . . . . . . . . . . . . 38
3.4 Toric Models and Exponential Matrices . . . . . . . . . . . . . . . . . . 43
4 Complex Projective Algebraic Statistics . . . . . . . . . . . . . . . . . . . . . 47
4.1 Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Projective Algebraic Models . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Parametric Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5 Conditional Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.1 Models of Conditional Independence . . . . . . . . . . . . . . . . . . . . 58
5.2 Markov Chains and Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3 Hidden Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

xv
xvi Contents

Part II Multi-linear Algebra

6 Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.1 Basic Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 The Tensor Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.3 Rank of Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.4 Tensors of Rank 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7 Symmetric Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.1 Generalities and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.2 The Rank of a Symmetric Tensor . . . . . . . . . . . . . . . . . . . . . . 106
7.3 Symmetric Tensors and Polynomials . . . . . . . . . . . . . . . . . . . . 110
7.4 The Complexity of Polynomials . . . . . . . . . . . . . . . . . . . . . . . . 113
7.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8 Marginalization and Flattenings . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
8.1 Marginalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
8.2 Contractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.3 Scan and Flattening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

Part III Commutative Algebra and Algebraic Geometry

9 Elements of Projective Algebraic Geometry . . . . . . . . . . . . . . . . . . 133
9.1 Projective Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
9.1.1 Associated Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
9.1.2 Topological Properties of Projective Varieties . . . . . . . 141
9.2 Multiprojective Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
9.3 Projective and Multiprojective Maps . . . . . . . . . . . . . . . . . . . . 148
9.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
10 Projective Maps and the Chow’s Theorem . . . . . . . . . . . . . . . . . . . 155
10.1 Linear Maps and Change of Coordinates . . . . . . . . . . . . . . . . . 155
10.2 Elimination Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.3 Forgetting a Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
10.4 Linear Projective and Multiprojective Maps . . . . . . . . . . . . . . . 163
10.5 The Veronese Map and the Segre Map . . . . . . . . . . . . . . . . . . 164
10.6 The Chow’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
10.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Contents xvii

11 Dimension Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

11.1 Complements on Irreducible Varieties . . . . . . . . . . . . . . . . . . . 179
11.2 Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
11.3 General Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
11.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
12 Secant Varieties . . . . . . . . . . . . . . . . . . . . . . . . ....... . . . . . . . . 195
12.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . ....... . . . . . . . . 195
12.2 Methods for Identifiability . . . . . . . . . . . . . ....... . . . . . . . . 203
12.2.1 Tangent Spaces and the Terracini’s Lemma . . . . . . . . . 203
12.2.2 Inverse Systems . . . . . . . . . . . . . . ....... . . . . . . . . 206
12.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . ....... . . . . . . . . 209
13 Groebner Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
13.1 Monomial Orderings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
13.2 Monomial Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
13.3 Groebner Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
13.4 Buchberger’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
13.5 Groebner Bases and Elimination Theory . . . . . . . . . . . . . . . . . 228
13.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
About the Authors

Prof. Cristiano Bocci is Assistant Professor of Geometry at the University of

Siena (Italy). His research concerns Algebraic Geometry, Commutative Algebra,
and their applications. In particular, his current interests are focused on symbolic
powers of ideals, Hadamard product of varieties, and the study of secant spaces. He
also works in two interdisciplinary teams in the fields of Electronic Measurements
and Sound Synthesis.

Prof. Luca Chiantini is Full Professor of Geometry at the University of Siena

(Italy). His research interests focus mainly on Algebraic Geometry and Multi-linear
Algebra, and include the theory of vector bundles on varieties and the study of
secant spaces, which are the geometric counterpart of the theory of tensor ranks. In
particular, he recently studied the relations between Multi-linear Algebra and the
theory of finite sets in projective spaces.

xix
 Part I
Algebraic Statistics
Chapter 1
Systems of Random Variables
and Distributions

1.1 Systems of Random Variables

This section contains the basic definitions with which we will construct our statistical
theory.
It is important to point out right away that in the field of Algebraic Statistics, a
still rapidly developing area of study, the basic definitions are not yet standardized.
Therefore, the definitions which we shall use in this text can differ significantly (more
in form than in substance) from those of other texts.
Definition 1.1.1 A random variable is a variable x taking values in a finite non-
empty set of symbols, denoted A(x). The set A(x) is called the alphabet of x or the
set of states of x. We will say that every element of A(x) is a state of the variable x.
A system of random variables (or random system) S is a finite set of random
variables.
The condition of finiteness, required both for the alphabet of a random variable
and the number of variables of a system, is typical of Algebraic Statistics. In other
statistical situations this hypothesis is often not present.
Definition 1.1.2 A subsystem of a system S of random variables is a system defined
by a subset S  ⊂ S.

Example 1.1.3 The simplest examples of a system of random variables are those
containing a single random variable. A typical example is obtained by thinking
of a die x as a random variable, i.e. as the unique element of S. Its alphabet is
A(x) = {1, 2, 3, 4, 5, 6}.
Another familiar example comes by thinking of the only element of S as a coin c
with alphabet A(c) = {H, T } (heads and tails).

Example 1.1.4 On internet sites about soccer betting one finds systems in which
each random variable has three states. More precisely the set S of random variables
are (say) all the professional soccer games in a given country. For each random
© Springer Nature Switzerland AG 2019 3
C. Bocci and L. Chiantini, An Introduction to Algebraic Statistics
with Tensors, UNITEXT 118, https://doi.org/10.1007/978-3-030-24624-2_1
4 1 Systems of Random Variables and Distributions

variable x, (i.e. game), its alphabet is A(x) = {1, 2, T }. The random variable takes
value “1” if the game was won by the home team, value “2” if the game was won by
the visiting team and value “T” if the game was a tie.

Example 1.1.5 (a) We can, similar to Example 1.1.3, construct a system S with two
random variables, namely with two dice {x1 , x2 }, both having alphabet A(xi ) =
{1, 2, 3, 4, 5, 6}.
(b) An example of another system of random variables T , closely related to
the previous one but different, is given by taking a single random variable
as the ordered pair of dice x = (x1 , x2 ) and, as alphabet A(x), all possible
values obtained by throwing the dice simultaneously: {(1, 1), . . . (1, 6), . . . ,
(6, 1), (6, 2), . . . , (6, 6)}.
(c) Another example W , still related to the two above (but different), is given by
taking as system the unique random variable the set consisting of two dice
z = {x1 , x2 } and as alphabet, A(z), the sum of the values of the two dice after
throwing them simultaneously: A(z) = {2, 3, 4, . . . , 12}.

Remark 1.1.6 The random variables of the systems S, T and W might seem, at first
glance, to be the same, but it is important to make clear that they are very different.
In (a) there are two random variables while in (b) and (c) there is only one. Also
notice that in T we have chosen an ordering of the two dice, while in W the random
variable is an unordered set of two dice. With example (a) there is nothing stopping
us from throwing the die x1 , say, twenty times and the die x2 ten times. However, in
both (b) and (c) the dice are each thrown the same number of times.

Example 1.1.7 There are many naturally occurring examples of systems with many
random variables. In fact, some of the most significant ones come from applications
in Economics and Biology and have an astronomical number of variables.
For example, in Economics and in market analysis, there are systems with one
random variable for each company which trades in a particular market. It is easy to
see that, in this case, we can have thousands, even tens of thousands, of variables.
In Biology, very important examples come from studying systems in which the
random variables represent hundreds (or thousands) of positions in the DNA sequence
of one or several species. The alphabet of each variable consists of the four basic
ingredients of DNA: Adenine, Cytosine, Guanine and Thymine. As a shorthand
notation, one usually denotes the alphabet of such random variables as {A, C, G, T }.
In this book, we will refer to the systems arising from DNA sequences, as DNA-
systems.

Example 1.1.8 For cultural reasons (one of the authors was born and lives in Siena!),
we will have several examples in the text of systems describing probabilistic events
related to the famous and colourful Sienese horse race called the Palio di Siena.
Horses which run in the Palio represent the various medieval neighbourhoods of the
city (called contrade) and the Palio is a substitute for the deadly feuds which existed
between the various sections of the city.
1.1 Systems of Random Variables 5

The names of the neighbourhoods are listed below with a shorthand letter abbre-
viation for each of them:

Aquila (eagle) (symbol: A) Bruco (caterpillar) (symbol: B)

Chiocciola (snail) (symbol: H) Civetta (little owl) (symbol: C)
Drago (dragon) (symbol: D) Giraffa (giraffe) (symbol: G)
Istrice (crested porcupine) (symbol: I) Leocorno (unicorn) (symbol: E)
Lupa (she-wolf) (symbol: L) Nicchio (conch) (symbol: N)
Oca (goose) (symbol: O) Onda (wave) (symbol: Q)
Pantera (panther) (symbol: P) Selva (forest) (symbol: S)
Tartuca (tortoise) (symbol: R) Torre (tower) (symbol: T)
Valdimontone (valley of the ram) (symbol: M).

Definition 1.1.9 A random variable x of a system S is called a boolean variable

if its alphabet has cardinality 2. A system is boolean if all its random variables are
boolean.
Remark 1.1.10 The states of a boolean random variable can be thought of as the
pair of conditions (true, false). As a matter of fact the standard alphabet of a boolean
random variable can be thought of as the elements of the finite field Z2 , where 1 =
true and 0 = false (this is our convention; be careful: in some texts this notation is
reversed!). Other alphabets, such as heads-tails or even-odd, are also often used for
the alphabets of boolean random variables.
Definition 1.1.11 A map or morphism between systems S and T of random variables
is a pair f = (F, G) where F is a function F : S → T and, for all x ∈ S, G defines
a function between alphabets G(x) : A(x) → A(F(x)).
The terminology used for functions can be transferred to maps of system of random
variables. Thus we can have injective maps (in which case both F and each of the
G(x) are injective), surjective maps (in which case both F and each of the G(x)
are surjective), isomorphism (in which case both F and all the maps G(x) are 1−1
correspondences). With respect to these definitions, the systems of random variables
form a category.
Example 1.1.12 If S  is a subsystem of S, the inclusion function S  → S defines, in
a obvious way, an injective map of systems. In this case, the maps between alphabets
are always represented by the identity map.
Example 1.1.13 Let S = {x} be the system defined by a die as in Example 1.1.3
with alphabet {1, 2, 3, 4, 5, 6}. Let T be the system defined by T = {y}, with
A(y) = {E, O} (E = even, O = odd). The function F : S → T , defined by F(x) =
y, and the function G : A(x) → A(y) defined by G(1) = G(3) = G(5) = O and
G(2) = G(4) = G(6) = E, define a surjective map from S to T of systems of ran-
dom variables.
The following definition will be of fundamental importance for the study of the
relationship between systems of random variables.
6 1 Systems of Random Variables and Distributions

Definition 1.1.14 The (total) correlation of a system S of random variables

{x1 , . . . , xn } is the system S = {x}, with a unique random variable x = (x1 , . . . , xn )
(the cartesian product of the elements x1 , . . . , xn of S). Its alphabet is given by
A(x1 ) × · · · × A(xn ), the cartesian product of the alphabets of the individual ran-
dom variables.

Remark 1.1.15 It is very important to notice that the definition of the total correlation
uses the concept of cartesian product. Moreover the concept of cartesian product
requires that we fix an ordering of the variables in S.
Thus, the total correlation of a system is not uniquely determined, but it changes
as the chosen ordering of the random variables changes.
It is easy to see, however, that all the possible total correlations of the system S
are isomorphic.

Example 1.1.16 If S is a system with two coins c1 , c2 , each having alphabet {H, T },
then the only random variable in its total correlation, has an alphabet with four
elements {(T, T ), (T, H ), (H, T ), (H, H )}, i.e. we have to distinguish between the
states (H, T ) and (T, H ). This is how the ordering of the coins enters into the
definition of the random variable (c1 , c2 ) of the total correlation.

Example 1.1.17 Let S be the system of random variables consisting of two dice, D1
and D2 each having alphabet the set {1, 2, 3, 4, 5, 6}. The total correlation of this
system, S, is the system with a unique random variable D = (D1 , D2 ) and alphabet
the set {(i, j) | 1 ≤ i, j ≤ 6}. So, the alphabet consists of 36 elements.
Now let T be the system whose unique random variable is the set x = {D1 , D2 }
and whose alphabet consists of the eleven numbers {2, 3, . . . , 11, 12}.
We can consider the surjective morphism of systems φ : S → T which takes
the unique random variable of S to the unique random variable of T and takes the
element (i, j) of the alphabet of the unique variable of S to i + j in the alphabet
of the unique variable of T .
Undoubtedly this morphism is familiar to us all!

Clearly if S is a system containing a single random variable, then X coincides

with its total correlation.

Definition 1.1.18 Let f : S → T be a map of systems of random variables, defined

by F : S → T and G(x) : A(x) → A(F(x)) for all random variables x ∈ S, and
suppose that F is bijective i.e. S and T have the same number of random variables.
Then f defines, in a natural way, a map f : S → T between the total corre-
lations as follows: for each state s = (s1 , . . . , sn ) of the unique variable (x1 , . . . , xn )
of S,
(f )(s) = (G(x1 )(s1 ), . . . , G(xn )(sn )).
1.2 Distributions 7

1.2 Distributions

One of the basic notions in the study of systems of random variables is the idea of a
distribution. Making the definition of a distribution precise will permit us to explain
clearly the idea of an observation on the random variables of a system. This latter
concept is extremely useful for the description of real phenomena.
Definition 1.2.1 Let K be any set. A K -distribution on a system S with random
variables x1 , . . . , xn , is a set of functions D = {D1 , . . . , Dn }, where for 1 ≤ i ≤ n,
Di is a function from A(xi ) to K .

Remark 1.2.2 In most concrete examples, K will be a numerical set, i.e. some subset
of C (the complex numbers).

The usual use of the idea of a distribution is to associate to each state of a variable
xi in the system S, the number of times (or the percentage of times) such a state is
observed in a sequence of observations.
Example 1.2.3 Let S be the system having as unique random variable a coin c, with
alphabet A(c) = {T, H } (the coin needs not be fair!).
Suppose we throw the coin n times and observe the state T exactly dT times and
the state H exactly d H times (dT + d H = n). We can use those observations to get an
N-distribution (N is the set of natural numbers), denoted Dc , where Dc : {T, H } → N
by
Dc (T ) = dT , Dc (H ) = d H .

One can identify this distribution with the element (dT , d H ) ∈ N2 .

We can define a different distribution, Dc on S (using the same series of observa-
tions) as follows:

Dc : {T, H } → Q (the rational numbers),

where Dc (T ) = dT /n and Dc (H ) = d H /n.

Example 1.2.4 Now consider the system S with two coins c1 , c2 , and with alphabets
A(ci ) = {T, H }.
Again, suppose we simultaneously throw both coins n times and observe that the
first coin comes up with state T exactly d1 times and with state H exactly e1 times,
while the second coin comes up T exactly d2 times and comes up H exactly e2 times.
From these observations we can define an N-distribution, D = (D1 , D2 ), on S
defined by the functions

D1 : {T, H } → N, D1 (T ) = d1 , D1 (H ) = e1 ,
D2 : {T, H } → N, D2 (T ) = d2 , D2 (H ) = e2
8 1 Systems of Random Variables and Distributions

It is also possible to identify this distribution with the element

((d1 , e1 ), (d2 , e2 )) ∈ N2 × N2 .

It is also possible to use this series of observations to define a distribution on the

total correlation S.
That system has a unique variable c = c1 × c2 = (c1 , c2 ) with alphabet A(c) =
{T T, T H, H T, H H }. The N-distribution on S we have in mind associates to each
of the four states how often that state appeared in the series of throws.
Notice that if we only had the first distribution we could not calculate the second
one since we would not have known (solely from the functions D1 and D2 ) how
often each of the states in the second system were observed.

Definition 1.2.5 The set of K -distributions of a system S of random variables forms

the space of distributions D K (S).

Example 1.2.6 Consider the DNA-system S with random variables precisely 100
fixed positions (or sites) p1 , . . . , p100 on the DNA strand of a given organism. As
usual, each variable has alphabet {A, C, G, T }. Since each alphabet has exactly four
members, the space of Z-distributions on S is D(S) = Z4 × · · · × Z4 (100times) =
Z400 .
Suppose we now collect 1,000 organisms and observe which DNA component
occurs in site i. With the data so obtained we can construct a Z-distribution D =
{D1 , . . . , D100 } on S where Di associates to each of the members of the alphabet
A( pi ) = {A, C, G, T } the number of occurrences of the corresponding component
in the i−th position. Note that for each Di we have

Di (A) + Di (C) + Di (G) + Di (T ) = 1,000.

Remark 1.2.7 Suppose that S is a system with random variables x1 , . . . , xn and that
the cardinality of each alphabet A(xi ) is exactly ai . As we have said before, ai is
simply the number of states that the random variable xi can assume.
With this notation, the K -distributions on S can be seen as points in the space

K a1 × · · · × K an .

We will often identify D K (S) with this space.

It is also certainly true that K a1 × · · · × K an = K a1 +···+an , and so it might seem
reasonable to say that this last is the set of distributions on S. However, since there
are so many ways to make this last identification, we could easily lose track of what
a particular distribution did on a member of the alphabet of one of the variables in S.

Remark 1.2.8 If S is a system with two variables x1 , x2 , whose alphabets have cardi-
nality (respectively) a1 and a2 , then the unique random variable in the total correlation
1.2 Distributions 9

S has a1 a2 states. Hence, as we said above, the space of K -distributions on S

could be identified with K a1 a2 .
Since we also wish to remember that the unique variable of S arises as the
cartesian product of the variables of S, it is even more convenient to think of
D K (S) = K a1 a2 as the set of a1 × a2 matrices with coefficients in K .
Thus, for a distribution D on S, we denote by Di j the value associated to the
state (i, j) of the unique variable, which corresponds to the states i of x1 and j of x2 .

For system with a bigger number of variables, we need to use multidimensional

matrices, commonly called tensors (see Definition 6.1.3).
The study of tensors is thus strongly connected to the study of systems of ran-
dom variables when we want to fix relationships among the variables (i.e. look at
distributions on the system). In fact, the algebra (and geometry) of spaces of tensors
represents the point of connection between the study of statistics on discrete sets and
other disciplines, such as Algebraic Geometry. The exploration of this connection is
our main goal in this book. We will take up that connection in another chapter.
Definition 1.2.9 Let S and T be two systems of random variables and f = (F, G) :
S → T a map of systems where F is a surjection. Let D be a distribution on S, and
D  a distribution on T .
The induced distribution f ∗D on T (called the image distribution) is defined as
follows: for t a state of the variable y ∈ T :

( f ∗D ) y (t) = Dx (s).
x∈F −1 (y),s∈G(x)−1 (t)

The induced distribution f D∗  on S (called the preimage distribution) is defined as

follows: for s a state of the variable x in S:

( f D∗  )x (s) = D F(x) (G(x)(s)).

We want to emphasize that distributions on a system of random variables should,

from a certain point of view, be considered as data on a problem. Data from which one
hopes to deduce other distributions or infer certain physical, biological or economic
facts about the system. We illustrate this idea with the following example.

Example 1.2.10 In the city of Siena (Italy) two spectacular horse races have been
run every year since the seventeenth century, with a few interruptions caused by the
World Wars. Each race is called a Palio, and the Palio takes place in the main square
of the city. In addition there have been some additional extraordinary Palios run from
time to time. From the last interruption, which ended in 1945, up to now (2014), a
total number of 152 Palios have taken place. Since the main square is large, but not
enormous, not every contrada can participate in every Palio. There is a method, partly
based on chance, that decides whether or not a contrada can participate in a particular
Palio.
10 1 Systems of Random Variables and Distributions

Table 1.1 Participation of the contrade at the 152 Palii (up to 2014)
x Name Dx (1) Dx (0) x Name Dx (1) Dx (0)
A Aquila 88 64 B Bruco 92 60
H Chiocciola 84 68 C Civetta 90 62
D Drago 95 57 G Giraffa 89 63
I Istrice 84 68 E Leocorno 99 52
L Lupa 89 63 N Nicchio 84 68
O Oca 87 65 Q Onda 84 68
P Pantera 96 56 S Selva 89 63
R Tartuca 91 61 T Torre 90 62
M Valdimontone 89 63

Let’s build a system with 17 boolean random variables, one for each contrada.
For each variable we consider the alphabet {0, 1}. The space of Z-distributions of
this system is Z2 × · · · × Z2 = Z34 .
Let us define a distribution by indicating, for each contrada x, Dx (1) = number
of Palios where contrada x took part and Dx (0) = number of Palios where contrada
x did not participate. Thus we must always have Dx (0) + Dx (1) = 152. The data
are given in Table 1.1.
We see that the Leocorno (unicorn) contrada participated in the most Palios while
the contrada Istrice (crested porcupine), Nicchio (conch), Onda (wave), Chiocciola
(snail) participated in the fewest.
On the same system, we can consider another distribution E, where E x (1) =
number of Palios that contrada x won and E x (0) = number of Palios that contrada x
lost (non-participation is considered a loss). The Win-Loss table is given in Table 1.2.
From the two tables we see that more participation in the Palios does not neces-
sarily imply more victories.

Table 1.2 Win-Loss table of contrade at the 152 Palii (up to 2014)
x Name E x (0) E x (1) x Name E x (0) E x (1)
A Aquila 8 144 B Bruco 5 147
H Chiocciola 9 143 C Civetta 8 144
D Drago 11 141 G Giraffa 12 140
I Istrice 8 144 E Leocorno 9 143
L Lupa 5 147 N Nicchio 9 143
O Oca 14 138 Q Onda 9 143
P Pantera 8 144 S Selva 15 137
R Tartuca 10 142 T Torre 3 149
M Valdimontone 9 143
1.3 Measurements on a Distribution 11

1.3 Measurements on a Distribution

We now introduce the concepts of sampling and scaling on a distribution for a system
of random variables.

Definition 1.3.1 Let K be a numerical set and let D = (D1 , . . . , Dn ) be a distribu-

tion on the system of random variables S = {x1 , . . . , xn }. The number

c D (xi ) = Di (s).
s∈A(xi )

is called the sampling of the variable xi in D. We will say that D has constant
sampling if all variables in S have the same sampling in D.
A K -distribution D on S is called probabilistic if each xi ∈ S has sampling equal
to 1.

Remark 1.3.2 Let S be a system with random variables {x1 , . . . , xn } and let D =
(D1 , . . . , Dn ) be a K -distribution on S, where K is a numerical field.
If every variable xi has sampling c D (xi ) = 0, we can obtain from D an associated
probabilistic distribution D̃ = ( D̃1 , . . . D̃n ) defined as follows:

Di (s)
for all i and for all states s ∈ A(xi ) set D̃i (s) = .
c D (xi )

Remark 1.3.3 In Example 1.2.3, the distribution D  is exactly the probabilistic dis-
tribution associated to D (seen as a Q-distribution).

Convention. To simplify the notation in what follows and since we will always be
thinking of the set K as some set of numbers, usually clear from the context, we
won’t mention K again but will speak simply of a distribution on a system S of
random variables.
Warning. We want to remind the reader again that the basic notation in Algebraic
Statistics is far from being standardized. In particular, the notation for a distribution
is quite varied in the literature and in other texts.
E.g. if si j is the j−th state of the i−th variable xi of the system S, and D is a
distribution on S, we will denote this by writing Di (si j ) as the value of D on that
state.
You will also find this number Di (si j ) denoted by Dxi =si j .

Example 1.3.4 Suppose we have a tennis tournament with 8 players where a player
is eliminated as soon as that player loses a match. So, in the first set of matches four
players are eliminated and in the second two more are eliminated and then we have
the final match between the remaining two players.
12 1 Systems of Random Variables and Distributions

We can associate to this tournament a system with 8 boolean random variables,

one variable for each player. We denote by D the distribution that, for each player
xi , is defined as:

Di (0) = number of matches lost;

Di (1) = number of matches won.

Clearly the sampling c(xi ) of every player xi represents the number of matches
played. For example, c(xi ) = 3 if and only if xi is a finalist, while c(xi ) = 1 for the
four players eliminated at the end of the first match. Hence D is not a distribution
with constant sampling.
Notice that this distribution doesn’t have any variable with sampling equal to 0
and hence there is an associated probabilistic distribution D̃, which represents the
statistics of victories. For example, for the winner xk , one has

D̃k (0) = 0, D̃k (1) = 1.

Instead, for a player x j eliminated in the semi-final,

1
D̃ j (0) = D̃ j (1) = .
2
While for a player xi eliminated after the first round we have

D̃i (0) = 1, D̃i (1) = 0.

The concept of an associated probabilistic distribution to a distribution D is quite

important in texts concerned with the analytic Theory of Probability. This is true to
such an extent that those texts work directly only with probabilistic distributions.
This is not the path we have chosen in this text. For us the concept that will be
more important than a probabilistic distribution is the concept of scaling. This latter
idea is more useful in connecting the space of distributions with the usual spaces in
which Algebraic Geometry is done.

Definition 1.3.5 Let D = (D1 , . . . , Dn ) be a distribution on a system S with random

variables {x1 , . . . , xn }. A distribution D  = (D1 , . . . , Dn ) is a scaling of D if, for any
x = xi ∈ S, there exists a constant λx ∈ K \ {0} such that, for all states s ∈ A(x),
Dx (s) = λx Dx (s).

Remark 1.3.6 Notice that the probabilistic distribution, D  , associated to a distribu-

tion D is an example of a scaling of D, where λx = 1/c(x).

Note moreover that, given a scaling D  of D, if D, D  have the same sampling, then
they must coincide.
1.3 Measurements on a Distribution 13

Remark 1.3.7 In the next chapters we will see that scaling doesn’t substantially
change a distribution. Using a projectivization method, we will consider two distri-
butions “equal” if they differ only by a scaling.
Proposition 1.3.8 Let f : S → T be a map of systems which is a bijection on the

sets of variables. Let D be a distribution on S and D  a scaling of D. Then f ∗D is a
scaling of f ∗D .
Proof Let y be a variable of T and let t ∈ A(y). Since f is a bijection there is a
unique x ∈ S for which f (x) = y. Then by definition we have

 
( f ∗D ) y (t) = D  (s) = λx D(s) = λx ( f ∗D ) y (t). 
s∈A(x) s∈A(x)

1.4 Exercises

Exercise 1 Let us consider the random system associated with the tennis tourna-
ment, see Example 1.3.4.
Compute the probabilistic distribution for the finalist who did not win the tourna-
ment.
Compute the probabilistic distribution for a tournament with 16 participants.
Exercise 2 Let S be a random system with variables x1 , . . . , xn and assume that all
the variables have the same alphabet A = {s1 , . . . , sm }. Then one can create the dual
system S  by taking s1 , . . . , sm as variables, each si with alphabet X = {x1 , . . . , xn }.
Determine the relation between the dimension of the spaces of K -distributions of
S and S  .
Exercise 3 Let S be a random system and let S  be a subsystem of S.
Determine the relation between the spaces of K -distributions of the correlations
of S and S  .
Exercise 4 Let f : S → S  be a surjective map of random systems.
Prove that if a distribution D on S  has constant sampling, then the same is true
for f D∗ .
Exercise 5 One can define a partial correlation over a system S, by connecting only
some of the variables.
For instance, if S has variables x1 , . . . , xn and m < n, one can consider the
partial correlation on the variables x1 , . . . , xm as a system T whose variables are
Y, xm+1 , . . . xn , where Y stands for the variable x1 × · · · × xm , with alphabet the
product A(x1 ) × · · · × A(xm ).
If S has variables c1 , c2 , c3 , all of them with alphabet {T, H } (see Example 1.1.3),
determine the space of K -distributions of the partial correlation T with random
variables c1 × c2 and c3 .
Another random document with
no related content on Scribd:
 But he, once satisfied—his restless and overweening ego
comforted by another victory—turning with a hectic and chronic, and
for him uncontrollable sense of satiety, as well as fear of
complications and burden—to other phases of beauty—other fields
and relationships where there was no such danger. For after all—
one more girl. One more experience. And not so greatly different
from others that had gone before it. And this in the face of the magic
of her meaning before capitulation. He did not understand it. He
could not. He did not even trouble to think about it much. But so it
was. And with no present consciousness or fear of being involved in
any early and unsatisfactory complications which might require
marriage—on the contrary a distinct and definite opposition to any
such complication at any time, anywhere.
Yet, at last, after many, many perfect hours throughout July and
August, the fatal complaint. There was something wrong, she feared.
She had such strange moods—such strange spells, pains, fears—
recently. Could there be? Did he think there could be any danger?
She had done what he said. Oh, if there was! What was she to do
then? Would he marry her? He must, really, then. There was no
other way. Her father—his fierce anger. Her own terrors. She could
not live at home any more. Could they not—would they not—be
married now if anything were wrong? He had said he would if
anything like this ever happened, had he not?
And Hauptwanger, in the face of this, suffering a nervous and cold
reaction. Marriage! The mere thought of such a thing! Impossible!
His father! His hitherto free roving life! His future! Besides, how did
she know? How could she be sure? And supposing she was! Other
girls got out of such things without much trouble. Why not she? And
had he not taken all the usual customary and necessary precautions
that he knew! She was too easily frightened—too uninformed—not
daring enough. He knew of lots of cases where girls got through
situations of this kind with ease. He would see about something first.
But conjoined with this, as she herself could see and feel, a
sudden definite coolness never before sensed or witnessed by her,
which was based on his firm determination not to pursue this
threatening relationship any longer, seeing that to do so meant only
to emphasize responsibility. And in addition, a keen desire to stay
away. Were there not other girls? A whole world full. And only
recently had he not been intrigued by one who was more aware of
the free, smart ways of pleasure and not so likely ever to prove a
burden?
But on the other hand, in the face of a father as strict as Zobel
himself and a mother who believed in his goodness, his course was
not absolutely clear either. And so from this hour on an attempt to
extricate himself as speedily and as gracefully as possible from this
threatening position. But before this a serious, if irritated, effort on his
part to find a remedy among his friends of the boating club and street
corners. But with the result merely of a vivid advertisement of the
fact that this gay and successful adventure of his had now resulted
most unsuccessfully for Ida. And thereafter hints and nods and
nudges among themselves whenever she chanced to pass. And Ida,
because of fear of scandal, staying in as much as she could these
days, or when she did appear trying to avoid Warren Avenue at High
as much as possible. For by now she was truly terrified, seized
indeed with the most weakening emotions based on the stern and
unrelenting countenance of her father which loomed so threateningly
beyond the immediate future. “If me no ifs,” and “but me no buts.”
Oh, how to do? For throughout the trial of this useless remedy, there
had been nothing to do but wait. And the waiting ended in nothing—
only greater horrors. And between all this, and enforced work at the
store and enforced duties at home, efforts to see her beloved—who,
because of new and more urgent duties, was finding it harder and
harder to meet her anywhere or at any time.
“But you must see how it is with me, don’t you, dear? I can’t go on
like this, can’t you see that? You said you’d marry me, didn’t you?
And look at all the time that’s gone already. Oh, I’m almost mad. You
must do something. You must! You must! If father should find out,
what in the world would I do? What would he do to me, and to you,
too? Can’t you see how bad it is?”
Yet in the face of this tortured plea on the part of this frantic and
still love-sick girl, a calm on the part of Hauptwanger that expressed
not indifference but cruelty. She be damned! He would not. He could
not. He must save himself now at whatever cost. And so a
determined attempt not to see her any more at all—never to speak to
her openly anywhere—or to admit any responsibility as to all this.
Yet, because of her inexperience, youth and faith thus far, no
willingness on her part to believe this. It could not be. She had not
even so much as sensed it before. Yet his continuing indifference
which could only be interpreted one way. The absences—the
excuses! And then one day, when pains and terror seized on her and
thereby drove her to him, he looking her calmly and brazenly in the
eye and announcing: “But I didn’t really promise to marry you, and
you know I didn’t. Besides, I’m not to blame any more than you are.
You don’t suppose that just because you don’t know how to take
care of yourself I’ve got to marry you now, do you?”
His eyes now for the first time were truly hard. His intention to end
this by one fell blow was very definite. And the blow was sufficient at
the moment to half unseat the romantic and all but febrile reason of
this girl, who up to this hour had believed so foolishly in love. Why,
how could this be? The horror of it! The implied disaster. And then
half in understanding, half in befuddled unreason, exclaiming: “But,
Ed! Ed! You can’t mean that. Why, it isn’t true! You know it isn’t! You
promised. You swore. You know I never wanted to—until you made
me. Why ... oh, what’ll I do now? My father! I don’t know what he’ll
do to me or to you either. Oh, dear! Oh, dear!” And frantically, and
without sufficient balance to warrant the name of reason, beginning
to wring her hands and twist and sway in a kind of physical as well
as mental agony.
At this Hauptwanger, more determined than ever to frighten her
away from him once and for all, if possible, exclaiming: “Oh, cut that
stuff! I never said I’d marry you, and you know it!” and turning on his
heel and leaving her to rejoin the chattering group of youths on the
corner, with whom, before her arrival, he had been talking. And as
much to sustain himself in this fatal decision as well as to carry it off
before them all, adding: “Gee, these skirts! It does beat hell, don’t
it?”
Yet now a little fearsome, if vain and contemptuous, for the
situation was beginning to take on a gloomy look. But just the same
when Johnny Martin, one of his companions and another aspirant for
street corner and Lothario honors, remarked: “I saw her here last
night lookin’ for you, Ed. Better look out. One of these skirts is likely
to do somepin to you one of these days”—he calmly extracted a
cigarette from a silver cigarette case and without a look in the
direction of the half-swooning Ida, said: “Is that so? Well, maybe.
We’ll see first.” And then with a nonchalant nod in the direction of
Ida, who, too tortured to even retreat, was standing quite still,
exclaimed: “Gee, these Germans! She’s got an old man that wouldn’t
ever let her find out anything and now because she thinks there’s
something wrong with her she blames me.” And just then, another
intimate approaching, and with news of two girls who were to meet
them somewhere later, exclaiming: “Hello, Skate! Everything set? All
right, then. We might as well go along. S’long, fellas.” And stepping
briskly and vigorously away.
But the stricken and shaken Ida still loitered under the already
partially denuded September trees. And with the speeding street and
auto cars with their horns and bells and the chattering voices and
shuffling feet of pedestrians and the blazing evening lights making a
kind of fanfare of color and sound. Was it cold? Or was it only herself
who was numb and cold? He would not marry her! He had never
said he would! How could he say that now? And her father to deal
with—and her physical condition to be considered!
As she stood there without moving, there flashed before her a
complete panorama of all the paths and benches of King Lake Park
—the little boats that slipped here and there under the trees at night
in the summertime—a boy and a girl—a boy and a girl—a boy and a
girl—to each boat. And the oars dragging most inconsequentially—
and infatuated heads together—infatuated hearts beating ecstatically
—suffocatingly strong. Yet now—after so many kisses and promises,
the lie given to her dreams, her words—his words on which her
words had been based—the lie given to kisses—hours, days, weeks,
months of unspeakable bliss—the lie given to her own security and
hopes, forever. Oh, it would be best to die—it would—it would.
And then a slow and dragging return to her room, where because
of the absence of her father and stepmother she managed to slip
into her bed and lie there, thinking. But with a kind of fever,
alternating with chills—and both shot through with most menacing
pains due to this most astounding revelation. And with a sudden and
keener volume of resentment than she had ever known gathering in
her brain. The cruelty! The cruelty! And the falsehood! He had not
only lied but insulted her as well. He who only five months before
had sought her so eagerly with his eyes and intriguing smile. The
liar! The brute! The monster! Yet linked and interwoven with such
thoughts as these, a lacerating desire not to believe them—to turn
back a month—two—three—to find in his eyes somewhere a trace of
something that would gainsay it all. Oh, Ed! Ed!
And so the night going—and the dawn coming. A horrible
lacerating day. And after that other days. And with no one to talk to—
no one. If only she could tell her stepmother all. And so other days
and nights—all alone. And with blazing, searing, whirling, disordered
thoughts in unbroken procession stalking her like demons. The
outside world in case she were to be thrust into it! Her own
unfamiliarity and hence fear of it! Those chattering, gaping youths on
the corners—the girls she knew—their thoughts, since they must all
soon know. Her loneliness without love. These and a hundred
related thoughts dancing a fantastic, macabre mental dance before
her.
But even so, within her own brain the persistent and growing
illusion that all she had heard from him was not true—a chimera—
and so for the time being at least continued faith in the value of
pleading. Her wonderful lover. It must be that still some
understanding could be reached. Yet with growing evidence that by
no plea or plaint was he to be restored to his former attitude. For, in
answer to notes, waiting at the corners, at the end of the street which
led down to his father’s coal-dock, in the vicinity of his home—
silence, evasions, or direct insults, and sneers, even.
“What’s the big idea, following me around, anyhow? You think I
haven’t anything else to do but listen to you? Say, I told you in the
first place I couldn’t marry, didn’t I? And now because you think
there’s something wrong with you, you want to make me
responsible. Well, I’m not the only fellow in this neighborhood. And
everybody knows that.”
He paused there, because as he saw this last declaration had
awakened in her a latent strength and determination never
previously shown in any way. The horror of that to her, as he could
see. The whiteness of her face afterwards and on the instant. The
blazing electric points within the pupils of her eyes. “That’s a lie, and
you know it! It’s not true! Oh, how terrible! And for you to say that to
me! I see it all now. You’re just a sneak and a coward. You were just
fooling with me all the time, then! You never intended to marry me,
and now because you’re afraid you think you can get out of it that
way—by trying to blame it on some one else. You coward! Oh, aren’t
you the small one, though! And after all the things you said to me—
the promises! As though I even thought of any one else in my life!
You dare to say that to me, when you know so well!”
Her face was still lily white. And her hands. Her eyes flashed with
transcendent and yet helpless and defeated misery. And yet, despite
her rage—in the center of this very misery—love itself—strong, vital,
burning love—the very core of it. But so tortured that already it was
beginning to drive the tears to her eyes.
And he knowing so thoroughly that this love was still there, now
instantly seizing on these latest truthful words of hers as an insult—
something on which to base an assumed grievance.
“Is that so? A coward, eh? Well, let’s see what you draw down for
that, you little dumb-bell.” And so turning on his heel—the strongest
instinct in him—his own social salvation in this immediate petty
neighborhood at the present time uppermost in his mind. And without
a look behind.
But Ida, her fear and terror at its height, calling: “Ed! Ed! You come
back here! Don’t you dare to leave me like this! I won’t stand for it. I
tell you, I won’t! You come back here now! Do you hear me?” And
seeing that he continued on briskly and indifferently, running after
him, unbelievably tense and a little beside herself—almost mentally
unaccountable for the moment. And he, seeing her thus and amazed
and troubled by this new turn his problem had taken, turning abruptly
with: “Say! You cut out o’ this now before I do something to you, do
you hear? I’m not the one to let you pull this stuff on me. You got
yourself into this and now you can get yourself out of it. Beat it
before I do something to you, do you hear?” And now he drew
nearer—and with such a threatening and savage look in his eyes
that for the first time in all her contact with him Ida grew fearful of
him. That angry, sullen face. Those fierce, cruel, savage eyes. Was it
really true that in addition to all the rest he would really do her
physical harm? Then she had not understood him at all, ever. And so
pausing and standing quite still, that same fear of physical force that
had kept her in subjection to her father overawing her here. At the
same time, Hauptwanger, noting the effect of his glowering rage,
now added: “Don’t come near me any more, do you hear? If you do,
you’re goin’ to get something you’re not going to like. I’m through,
and I’m through for good, see.”
Once more he turned and strode away, this time toward the central
business district of the High and Warren Avenue region—the while
Ida, too shaken by this newest development to quite grasp the full
measure of her own necessity or courage, stood there. The horror of
it! The disgrace! The shame! For now, surely, tragedy was upon her!
For the time being, in order to save herself from too much
publicity, she began to move on—walk—only slowly and with
whirling, staggering thoughts that caused her to all but lurch. And so,
shaking and pale, she made her way once more to her home, where
she stole into her room unnoticed. Yet, now, too tortured to cry but
thinking grimly—fiercely at moments—at other times most weakly
and feebly even—on all that had so recently occurred.
Her father! Her stepmother! If he—she—they should come to
know! But no—something else must happen before ever that should
be allowed to happen. She must leave—or—or, better yet—maybe
drown herself—make way with herself in some way—or—
In the garret of this home, to which as a child on certain days she
had frequently resorted to play, was an old wire clothes-line on which
was hung an occasional wash. And now—might not that—in the face
of absolute fiasco here—might not that—she had read of ending
one’s life in that manner. And it was so unlikely that any one would
trouble to look there—until—until—well—
But would she? Could she? This strange budding of life that she
sensed—feared. Was it fair to it? Herself? To life that had given it to
her? And when she desired so to live? And when he owed her
something—at least help to her and her—her—her—No, she could
not—would not think of that yet, especially when to die this way
would be but to clear the way to easier and happier conquests for
him. Never! Never! She would kill him first—and then herself. Or
expose him and so herself—and then—and then—
But again her father! Her stepmother! The disgrace! And so—
In her father’s desk at the store was a revolver—a large, firm,
squarish mechanism which, as she had heard him say, fired eight
shots. It was so heavy, so blue, so cold. She had seen it, touched it,
lifted it once—but with a kind of terror, really. It was always so
identified with death—anger—not life—But now—supposing—
supposing, if she desired to punish Edward and herself—or just
herself alone. But no, that was not the way. What was the way,
anyhow? What was the way?
And so now brooding in a tortured and half-demented way until her
father, noting her mental state, inquired solemnly as to what had
come over her of late. Had she had a quarrel with Hauptwanger? He
had not seen him about recently. Was she ill in any way? Her
appetite had certainly fallen off. She ate scarcely anything. But
receiving a prompt “No” to both inquiries he remained curious but
inclined to suspend further inquiry for the time being. There was
something, of course, but no doubt it would soon come out.
But now—in the face of this—of course there must be action—
decision. And so, in view of the thoughts as to self-destruction and
the revolver, a decision to try the effect of a physical threat upon
Hauptwanger. She would just frighten him. She might even point the
gun at him—and see what he would do then. Of course, she could
not kill him—she knew that. But supposing—supposing—one aimed
—but not at him, really—and—and—(but, oh no!) a spit of fire, a puff
of smoke, a deadly bullet—into his heart—into hers afterward, of
course. No, no! For then what? Where?
A dozen, a score, of times in less than two days she approached
the drawer that held the revolver and looked at it—finally lifting it up
but with no thought of doing more than just that at the moment. It
was so heavy, so cold, so blue. The very weight and meaning of it
terrorized her, although at last—after the twentieth attempt—she was
able to fit it into her bosom in such a way that it lay quite firm and
still. The horror of it—cold against her breast, where so often during
the summer his head had lain.
And then one afternoon, when she could scarcely endure the
strain longer—her father demanding: “What is the matter with you,
anyhow? Do you know what you’re doing half the time? Is there
anything wrong between you and that beau of yours? I see he
doesn’t come around any more. It is time that you either married or
had nothing more to do with him, anyway. I don’t want any silly
nonsense between you and him, you know.” And this effected the
very decision which she had most dreaded. Now ... now ... she must
act. This evening—at least she must see him again and tell him that
she was going to see his father and reveal all—furthermore, that if
he did not marry her she would kill him and herself. Show him the
gun, maybe, and frighten him with it—if she could—but at any rate
make a last plea as well as a threat. If only—if only he would listen
this time—not turn on her—become frightened, maybe, and help her,
—not curse—or drive her away.
There was the coal-yard of his father that was at the end of an
inlet giving into the river. Or his own home. She might go first to the
coal office. He would be sure to be leaving there at half past five, or
at six he would be nearing his home. At seven or half past departing
from that again very likely to see—to see—whom? But best—best to
go to the coal office first. He would be coming from there alone. It
would be the quickest.
And Hauptwanger coming out of the coal office on this particular
evening in the mood and with the air of one with whom all was well.
But in the windy dusk of this November evening, arc lights blazing in
the distance, the sound of distant cars, distant life, the wind whipping
crisp leaves along the ground—the figure of a girl—a familiar cape
about her shoulders, suddenly emerging from behind a pile of brick
he was accustomed to pass.
“Ed! I want to talk to you a minute.”
“You again! What the hell did I tell you? I ain’t got no time to talk to
you, and I won’t! What did I....”
“Now listen, Ed, stop that, now! I’m desperate. I’m desperate, Ed,
do you hear? Can’t you see?” Her voice was staccato—almost shrill
and yet mournful, too. “I’ve come to tell you that you’ve got to marry
me now. You’ve got to—do you hear?” She was fumbling at her
breast where lay that heavy blue thing—no longer so cold as when
she had placed it there. The handle was upward. She must draw it
now—show it—or hold it under her cloak ready so that at the right
moment she could show it—and make him understand that unless
he did something.... But her hand shaking so that she could scarcely
hold it. It was so heavy—so terrible. She could scarcely hear herself
adding: “Otherwise, I’m going to your father and mine, now. My
father may do something terrible to me but he’ll do more to you. And
so will your father when he knows.... But, anyway....” She was about
to add: “You’ve got to marry me, and right away too, or, or, I’m going
to kill you and myself, that’s all—” and then to produce the revolver,
and wave it before him in a threatening dramatic manner.
But before that the uncalculated and non-understanding fury of
Hauptwanger. “Well, of all the nerve! Say, cut this out, will ya? Who
do you think you are? What did I tell you? Go to my father, if you
want to. Go to yours! Who’s afraid? Do you think they’re going to
believe a —— like you? I never had anything to do with you, and
that’s that!” And then in his anger giving her a push—as much to
overawe her as anything.
But then, in spite of her desire not to give way, fury, blindness,
pain,—whirling, fiery sparks, such as never in all her life before had
she seen—and executing strange, rhythmic, convoluting orbits in her
brain—swift, eccentric, red and yet beautiful orbits. And in the center
of them the face of Hauptwanger—her beloved—but not as it was
now—oh, no—but rather haloed by a strange white light—as it was
under the trees in the spring. And herself turning, and in spite of the
push, jumping before him.
“You will marry me, Ed, you will! You will! You see this? You will
marry me!”
And then, as much to her astonishment as to his—yet with no
particular terror to either of them—the thing spitting flame—making a
loud noise—jumping almost out of her hand—so much so that before
she could turn it away again there was another report—another flash
of red in the dusk. And then Hauptwanger, too astonished quite for
words at the moment, exclaiming: “Jesus! What are you....” And
then, because of a sharp pain in his chest, putting his hand there
and adding: “Oh, Christ! I’m shot!” and falling forward to one side of
her....
And then herself—those same whirling red sparks in her brain,
saying: “Now, now—I must kill myself, too. I must. I must. I must run
somewhere and turn this on myself,” only quite unable to lift it at the
moment—and because of some one—a man—approaching—a
voice—footsteps, running—herself beginning to run—for some tree
—some wall—some gate or doorway where she might stop and fire
on herself. But a voice: “Hey! Stop that girl!” “Murder!” And another
voice from somewhere else: “Hey! Murder! Stop that girl!” And
footsteps, hard, quick ones, immediately behind her. And a hand
grabbing hers in which was still the pistol, wildly and yet unwittingly
held. And as the other hand wrenched at her hand—“Gimme that
gun!” And then a strong youth whom she had never seen before—
and yet not unlike Eddie either—turning her about—restraining her—
“Say, you! What the devil is this, anyhow? Come back here. You
can’t get away with this.”
And yet at the same time not unfriendly eyes looking into hers,
strong hands holding her, but not too roughly, and herself exclaiming:
“Oh, let me go! Let me go! I want to die, too, I tell you! Let me go!”
And sobbing great, dry, shaking sobs.
But after that—and all so quickly—crowds—crowds—men and
women, boys and girls, and finally policemen gathering about her,
each with the rules of his training firmly in mind to get as much
general information as possible; to see that the wounded man was
hurried to a hospital, the girl to a precinct police station; the names
and addresses of various witnesses secured. But with the lorn Ida in
a state of collapse—seated upon a doorstep in a yard surrounded by
a pushing crowd, while voices rang in her ears: “Where? What?
How?” “Sure, sure! Just now, right back there. Sure, they’re calling
the ambulance.” “He’s done for, I guess. Twice in the breast. He
can’t live.” “Gee! He’s all covered with blood.” “Sure, she did. With a
revolver—a great big one. The cop’s got it. She was tryin’ to get
away. Sure, Jimmie Allen caught her. He was just comin’ home.”
“Yeah. She’s the daughter of old Zobel who keeps the paint store up
here in Warren Avenue. I know her. An’ he’s the son of this
Hauptwanger here who owns the coal-yard. I used to work for ’em.
He lives up in Grey Street.”
But in the meantime young Hauptwanger unconscious and being
transferred to an operating table at Mercy Hospital—his case
pronounced hopeless—twenty-four hours of life at the very most.
And his father and mother hearing the news and running there. And
in the same period the tortured Ida transferred to Henderson Avenue
Police Station, where in a rear inquisitorial chamber, entirely
surrounded by policemen and detectives, she was questioned and
requestioned. “Yah say yah seen this fella for the first time over a
year ago? Is that right? He just moved into the neighborhood a little
while before? Ain’t that so?” And the disconsolate, half-conscious Ida
nodding her head. And outside a large, morbid, curious crowd. A
beautiful girl! A young man dying! Some sex mystery here.
And in the interim Zobel himself and his wife, duly informed by a
burly policeman, hurrying white-faced and strained to the station. My
God! My God! And both rushing in breathless. And beads of
perspiration on Zobel’s forehead and hands—and misery, misery
eating at his vitals. What! His Ida had shot some one! Young
Hauptwanger! And in the street, near his office! Murder! Great God!
Then there was something between them. There was. There was.
But might he not have known? Her white face. Her dreary, forsaken
manner these later days. She had been betrayed. That was it.
Devils! Devils! That was it! Eighty thousand hells! And after all he
had said to her! And all his and his wife’s care of her! And now the
neighbors! His business! The police! A public trial! Possibly a
sentence—a death sentence! God in heaven! His own daughter, too!
And that young scoundrel with his fine airs and fine clothes! Why—
why was it that he had let her go with him in the first place? When he
might have known—his daughter so inexperienced. “Where is she?
My God! My God! This is terrible!”
But seeing her sitting there, white, doleful-looking, and looking up
at him when spoken to with an almost meaningless look—a
bloodless, smileless face—and saying: “Yes, I shot him. Yes. Yes. He
wouldn’t marry me. He should have but he wouldn’t—and so—” And
then at once crushing her hands in a sad, tortured way and crying:
“Oh, Ed! Ed!” And Zobel exclaiming: “Ach, God! Ida! Ida! In God’s
name, it can’t be so. Why didn’t you tell me? Why didn’t you come to
me? Am I not your father! I would have understood. Of course! Of
course! I would have gone to his father—to him. But now—this—and
now—” and he began to wring his own hands.
Yet the principal thought in his mind that now the world would
know all— And after all his efforts. And beginning volubly to explain
to the desk lieutenant and the detectives and policemen all that he
knew. But the only thought afloat in the unhappy Ida’s brain, once
she awakened again, was: Was this really her father? And was he
talking so—of help? That she might have come to him—for what—
when she had thought—that—that he would not be like that to her.
But ... after a time again ... there was Ed to be thought of. That
terrible scene. That terrible accident. She had not intended to do that
—really. She had not. She had not. No! But was he really dead? Had
she really killed him? That push—almost a blow it was—those
words. But still— Oh, dear! Oh, dear! And then beginning to cry to
herself, silently and deeply, while Zobel and his wife bent over her for
the first time in true sympathy. The complications of life! The terrors!
There was no peace for any one on this earth—no peace—no
peace. All was madness, really, and sorrow. But they would stand by
her now—yes, yes.
 But then the reporters. A public furore fanned by the newspapers,
with their men and women writers, pen and ink artists,
photographers. Their editorials: “Beautiful girl of seventeen shoots
lover, twenty-one. Fires two bullets into body of man she charges
with refusing to keep faith. About to become a mother. Youth likely to
die. Girl admits crime. Pleads to be left alone in misery. Parents of
both in despair.” And then columns and columns, day after day—
since on the following afternoon at three Hauptwanger did die—
admitting that he had wronged her. And a coroner’s jury, called
immediately afterwards, holding the girl for subsequent action by the
Grand Jury, and without bail. Yet, because of her beauty and the
“pathos” of the case—letters to the newspapers, from ministers,
society men and women, politicians and the general public,
demanding that this wronged girl about to become a mother, and
who had committed no wrong other than that of loving too well—if
not wisely—be not severely dealt with—be forgiven—be admitted to
bail. No jury anywhere would convict her. Not in America. Indeed, it
would “go hard” with any jury that would attempt to “further punish” a
girl who had already suffered so much. Plainly it was the duty of the
judge in this case to admit this poor wronged soul to bail and the
peace and quiet of some home or institution where her child might
be born, especially since already a woman of extreme wealth and
social position, deeply stirred by the pathos of this drama, had not
only come forward to sympathize with this innocent victim of love
and order and duty, but had offered any amount of bail that she
might be released to the peace and quiet of her own home—there to
await the outcome of her physical condition as well as the
unavoidable prosecution which must fix her future.
And so, to her wonder and confusion, Ida finally released in the
custody of this outwardly sober and yet inwardly emotional woman,
who ever since the first day of her imprisonment in the central county
jail had sought to ingratiate herself into her good graces and
emotions—a woman middle-aged and plain but soft-voiced and
kindly-mannered, who over and over repeated that she understood,
that she also had suffered—that her heart had been torn, too—and
that she, Ida, need never worry. And so Ida finally transferred (a
bailed prisoner subject to return upon demand) to the wide acres and
impressive chambers of a once country but now city residence, an
integral part of the best residence area of the city. And there, to her
astonishment and wonder—and this in spite of her despair—all
needful equipment and service provided—a maid and servants, her
food served to her in her room when she wished—silence or
entertainment as she chose. And with her own parents allowed to
visit her whenever she chose. Yet she was so uncomfortable in their
presence always now. True, they were kind—gentle, whenever they
came. They spoke of the different life that was to be after this great
crisis was truly past—the birth of the child, which was never other
than indirectly referred to, or the trial, which was to follow later. There
was to be a new store in a new neighborhood. The old one had
already been offered for sale. And after that ... well, peace perhaps,
or a better life. But even in her father’s eyes as he spoke could she
not see the weight of care which he now shouldered? She had
sinned! She had killed a man! And wrecked another family—the
hearts of two other parents as well as her father’s own peace of mind
and commercial and social well-being. And in all his charity, was
there room for that? In the solemnity of his manner, as well as that of
her stepmother, could she feel that there was?
Yet in the main, and because her mood and health seemed to
require it, she was now left to contemplate the inexplicable chain of
events which her primary desire for love had brought about. The
almost amazing difference in the mental attitude of her parents
toward her now and before this dreadful and unfortunate event in her
life! So considerate and sympathetic now as to result in an offer of a
happier home for her and her child in the future, whereas before all
was—or as she sensed it—so threatening and desperate. The
strange and to her inexplicable attitude of this woman even—so kind
and generous—and this in the face of her sin and shame.
And yet, what peace or quiet could there be for her here or
anywhere now? The terrible torture that had preceded that terrible
accident! Her Edward’s cry! His death! And when she loved him so!
Had! Did now! And yet by his dread perverseness, cruelty, brutality,
he had taken himself from her. But still, still—now that he was gone
—now that in dying, as she heard he had said, he had been “stuck
on her” at first, that she had “set him crazy,” but that afterwards,
because of his parents, as well as hers, he had decided that he
would not marry her—she could not help but feel more kindly to him.
He had been cruel. But had he not died? And at her hands. She had
killed him—murdered him. Oh, yes, she had. Oh! Oh! Oh! For in
connection with the actual scene did she not recall some one crying
that his shirt front had been all bloody. Oh! Oh! Oh! And in her heart,
no doubt, when she had jumped in front of him there in the dusk had
been rage—rage and hate even, too, for the moment. Oh, yes. But
he had cried: “Oh, Christ! I’m shot!” (Her Edward’s cry.)
And so, even in the silence of these richly furnished rooms, with a
servant coming to her call, hot, silent tears and deep, racking sobs—
when no one was supposed to see or hear—and thoughts, thoughts,
thoughts—sombre, bleak—as to her lack of sense, her lack of
courage or will to end it all for herself on that dreadful evening when
she so easily might have. And now here she had plighted her word
that she would do nothing rash—would not attempt to take her life.
But the future! The future! And what had she not seen since that
dreadful night! Edward’s father and mother at the inquest! And how
they had looked at her! Hauptwanger, senior—his strong, broad
German face marked with a great anguish. And Mrs. Hauptwanger—
small and all in black, and with great hollow rings under her eyes.
And crying silently nearly all the time. And both had sworn that they
knew nothing of Edward’s conduct, or of his definite interest in her.
He was a headstrong, virile, restless boy. They had a hard time
controlling him. And yet he had not been a bad boy, either—
headstrong but willing to work—and gay—their only son.
At one point in these extensive grounds—entirely surrounded by
Lombardy poplars now leafless—there stood a fountain drained of its
water for the winter. But upon the pedestal, upon a bronze rock, at
the foot of which washed bronze waves of the Rhine, a Rhine
maiden of the blonde German Lorelei type, standing erect and a-
dream, in youth, in love. And at her feet, on his knees, a German
lover of the Ritter type—vigorous, uniformed, his fair blond head and
face turned upward to the beauty about whose hips his arms were
clasped—his look seeking, urgent. And upon his fair bronze hair, her
right hand, the while she bent on him a yearning, yielding glance.
Oh, Edward! Oh, love! Spring! She must not come out here any
more. And yet evening after evening in early December, once the
first great gust of this terrific storm had subsided and she was seeing
things in a less drastic light, she was accustomed to return to look at
it. And sometimes, even in January, a new moon overhead would
suggest King Lake Park! The little boats gliding here and there! She
and Edward in one. Herself leaning back and dreaming as now—
now—this figure of the girl on the rock was doing. And he—he—at
her knees. To be sure, he had cursed her. He had said the
indifferent, cruel words that had at last driven her to madness. But
once he had loved her just this way. It was there, and only there, that
she found spiritual comfort in her sorrow—
But then, in due course, the child—with all these thoughts, moods
enveloping it. And after that the trial, with her prompt acquittal. A
foreseen conclusion. And with loud public acclaim for that verdict
also, since it was all for romance and drastic drama. And then the
final leaving of these great rooms and this personal intimate affection
that had been showered upon her. For after all the legal, if not the
emotional problem, had now been solved. And since her father was
not one who was poor or welcomed charity—a contemplated and
finally accomplished return to a new world—the new home and store
which had been established in a very different and remote part of the
city. The child a boy. That was good, for eventually he could care for
himself. He would not need her. The new paint shop was near
another cross business street, near another moving-picture theatre.
And boys and girls here as elsewhere—on the corners—going arm
in arm—and herself again at home cooking, sewing, cleaning as
before. And with Mrs. Zobel as reserved and dubious as before. For
after all, had she not made a mess of her life, and for what? What
now? Here forever as a fixture? And even though Zobel, in spite of
his grimness, was becoming fond of the child. How wretched, how
feeble life really was!
But far away King Lake Park and the old neighborhood. And
thoughts that went back to it constantly. She had been so happy the
summer before. And now this summer! And other summers to come
—even though perhaps some time—once little Eric was grown—
there might be some other lover—who would not mind— But, no—
no, not that. Never! She did not want that. Could not—would not
endure it.
And so at last of a Saturday afternoon, when she had the excuse
of certain things needed for Eric, a trip to presumably the central
business heart—whereas, in reality, it was to King Lake Park she
was going. And once there—the little boats, the familiar paths—a
certain nook under the overarching bushes and trees. She knew it so
well. It was here that she had demanded to be let out in order that
she might go home by herself—so shocked, so ashamed. Yet now
seeking it.
The world does not understand such things. It is so busy with so
many, many things.
And then dusk—though she should have been returning. Her boy!
He would miss her! And then a little wind with a last faint russet glow
in the west. And then stars! Quite all the world had gone to its dinner
now. The park was all but empty. The water here was so still—so
agate. (The world—the world—it will never understand, will it?)
Where would Edward be? Would he be meeting her somewhere?
Greeting her? Would he forgive,—when she told him all—could she
find him, perchance? (The world—the busy, strident, indifferent,
matter-of-fact world—how little it knows.)
And then a girl in the silence, in the shadow, making her way down
to the very spot that the nose of their boat had nuzzled but one short
summer before. And calmly stepping into the water and wading out
to her knees—to her waist—her breasts—in the mild, caressing
water—and then to her lips and over them—and finally, deliberately
—conclusively—sinking beneath its surface and without a cry or
sigh.
The world does not understand such things. The tide of life runs
too fast. So much that is beautiful—terrible—sweeps by—by—by—
without thought—without notice in the great volume.
 And yet her body was found—her story retold in great, flaring
headlines. (Ida Zobel—Girl Slayer of Hauptwanger a Suicide.) And
then ... and then ... forgotten.
 VIII
THE OLD NEIGHBORHOOD

H E came to it across the new bridge, from the south where the
greater city lay—the older portion—and where he had left his
car, and paused at the nearer bridgehead to look at it—the eddying
water of the river below, the new docks and piers built on either side
since he had left, twenty years before; the once grassy slopes on the
farther shore, now almost completely covered with factories,
although he could see too, among them, even now, traces of the old,
out-of-the-way suburb which he and Marie had known. Chadds
Bridge, now an integral part of the greater city, connected by car
lines and through streets, was then such a simple, unpretentious
affair, a little suburban village just on the edge of this stream and
beyond the last straggling northward streets of the great city below,
where the car lines stopped and from which one had to walk on foot
across this bridge in order to take advantage of the rural quiet and
the cheaper—much cheaper—rents, so all-important to him then.
Then he was so poor—he and Marie—a mere stripling of a
mechanic and inventor, a student of aeronautics, electricity,
engineering, and what not, but newly married and without a dollar,
and no clear conception of how his future was to eventuate, whereas
now—but somehow he did not want to think of now. Now he was so
very rich, comparatively speaking, older, wiser, such a forceful
person commercially and in every other way, whereas then he was
so lean and pathetic and worried and wistful—a mere uncertain
stripling, as he saw himself now, with ideas and ambitions and
dreams which were quite out of accord with his immediate prospects
or opportunities. It was all right to say, as some one had—Emerson,
he believed—“hitch your wagon to a star.” But some people hitched,
or tried to, before they were ready. They neglected some of the
slower moving vehicles about them, and so did not get on at all—or
did not seem to, for the time being.