Object Oriented Data Analysis 1st Edition Marron

pdf download

https://ebookmeta.com/product/object-oriented-data-analysis-1st-
edition-marron/

Download more ebook from https://ebookmeta.com

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

Object Oriented Data Analysis 1st Edition James Stephen

Marron

https://ebookmeta.com/product/object-oriented-data-analysis-1st-
edition-james-stephen-marron/

Object-Oriented Python Irv Kalb

https://ebookmeta.com/product/object-oriented-python-irv-kalb/

Object-Oriented Python 1st Edition Irv Kalb

https://ebookmeta.com/product/object-oriented-python-1st-edition-
irv-kalb/

Genome Editing Current Technology Advances and

Applications for Crop Improvement 1st Edition Shabir
Hussain Wani

https://ebookmeta.com/product/genome-editing-current-technology-
advances-and-applications-for-crop-improvement-1st-edition-
shabir-hussain-wani/
Summary Analysis of White Fragility Why It s So Hard
for White People to Talk About Racism A Guide to the
Book by Robin DiAngelo 1st Edition Zip Reads

https://ebookmeta.com/product/summary-analysis-of-white-
fragility-why-it-s-so-hard-for-white-people-to-talk-about-racism-
a-guide-to-the-book-by-robin-diangelo-1st-edition-zip-reads/

Chimpanzees and Human Evolution 1st Edition Martin N.

Muller

https://ebookmeta.com/product/chimpanzees-and-human-
evolution-1st-edition-martin-n-muller/

Immunology of Endometriosis: Pathogenesis and

Management 1st Edition Kaori Koga

https://ebookmeta.com/product/immunology-of-endometriosis-
pathogenesis-and-management-1st-edition-kaori-koga/

Cultures of Change in Contemporary Zimbabwe Socio

Political Transition from Mugabe to Mnangagwa Routledge
Contemporary Africa 9781032040264 1st Edition Oliver
Nyambi (Editor)
https://ebookmeta.com/product/cultures-of-change-in-contemporary-
zimbabwe-socio-political-transition-from-mugabe-to-mnangagwa-
routledge-contemporary-africa-9781032040264-1st-edition-oliver-
nyambi-editor/

Landscaping for Dummies 2nd Edition Teri Dunn Chace

https://ebookmeta.com/product/landscaping-for-dummies-2nd-
edition-teri-dunn-chace/
Tu rkiyenin Tarihi Deniz Fenerleri Historical
Lighthouses of Turkey First Edition Bostan ■dris
Türkhan M Sait Soyer Kolçak Emel

https://ebookmeta.com/product/tu-rkiyenin-tarihi-deniz-fenerleri-
historical-lighthouses-of-turkey-first-edition-bostan-idris-
turkhan-m-sait-soyer-kolcak-emel/
Object Oriented
Data Analysis
MONOGRAPHS ON STATISTICS AND APPLIED
PROBABILITY

Editors: F. Bunea, R. Henderson, N. Keiding, L. Levina, R. Smith, W.

Wong

Recently Published Titles

Large Covariance and Autocovariance Matrices

Arup Bose and Monika Bhattacharjee 162

The Statistical Analysis of Multivariate Failure Time Data: A

Marginal Modeling Approach
Ross L. Prentice and Shanshan Zhao 163

Dynamic Treatment Regimes

Statistical Methods for Precision Medicine
Anastasios A. Tsiatis, Marie Davidian, Shannon T. Holloway, and Eric
B. Laber 164

Sequential Change Detection and Hypothesis Testing

General Non-i.i.d. Stochastic Models and Asymptotically Optimal Rules
Alexander Tartakovsky 165

Introduction to Time Series Modeling

Genshiro Kitigawa 166

Replication and Evidence Factors in Observational Studies

Paul R. Rosenbaum 167

Introduction to High-Dimensional Statistics, Second Edition

Christophe Giraud 168

Object Oriented Data Analysis

J.S. Marron and Ian L. Dryden 169

Martingale Methods in Statistics

Yoichi Nishiyama 170

For more information about this series please visit: https://www.crcpress.com/

Chapman–HallCRC-Monographs-on-Statistics–Applied-Probability/book-series/
CHMONSTAAPP
Object Oriented
Data Analysis

J.S. Marron and Ian L. Dryden

First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

© 2022 Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and
let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.
copyright.com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. For works that are not available on CCC please contact
mpkbookspermissions@tandf.co.uk

Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Marron, James Stephen, 1954- author. | Dryden, I. L. (Ian L.),
author.
Title: Object oriented data analysis / J.S. Marron and Ian L. Dryden.
Description: [Boca Raton] : Taylor & Francis Group, LLC, [2021] | Includes
bibliographical references and index. | Summary: “Object Oriented Data
Analysis (OODA) provides a useful general framework for the
consideration of many types of Complex Data. It is deliberately intended
to be particularly useful in the analysis of data in complicated
situations which are typically not easily represented as an
unconstrained matrix of numbers”-- Provided by publisher.
Identifiers: LCCN 2021023347 (print) | LCCN 2021023348 (ebook) | ISBN
9780815392828 (hardback) | ISBN 9781032114804 (paperback) | ISBN
9781351189675 (ebook)
Subjects: LCSH: Object-oriented methods (Computer science) | Quantitative
research. | Statistics--Methodology.
Classification: LCC QA76.9.O35 M369 2021 (print) | LCC QA76.9.O35 (ebook)
| DDC 005.1/17--dc23
LC record available at https://lccn.loc.gov/2021023347
LC ebook record available at https://lccn.loc.gov/2021023348

ISBN: 978-0-8153-9282-8 (hbk)

ISBN: 978-1-032-11480-4 (pbk)
ISBN: 978-1-351-18967-5 (ebk)

DOI: 10.1201/9781351189675

Typeset in Nimbus
by KnowledgeWorks Global Ltd.
 Dedication

To our families for their ongoing strong support over the many years it took to
fully develop these ideas, and to the many colleagues who have played a vital role
in shaping this approach to data analysis.

v
 Contents

Preface xi

1 What Is OODA? 1
1.1 Case Study: Curves as Data Objects 3
1.2 Case Study: Shapes as Data Objects 10
1.2.1 The Segmentation Challenge 10
1.2.2 General Shape Representations 12
1.2.3 Skeletal Shape Representations 13
1.2.4 Bayes Segmentation via Principal Geodesic Analysis 15

2 Breadth of OODA 19
2.1 Amplitude and Phase Data Objects 19
2.2 Tree-Structured Data Objects 23
2.3 Sounds as Data Objects 25
2.4 Images as Data Objects 28

3 Data Object Definition 31

3.1 OODA Foundations 31
3.1.1 OODA Terminology 31
3.1.2 Object and Feature Space Example 32
3.1.3 Scree Plots 36
3.1.4 Formalization of Modes of Variation 38
3.2 Mathematical Notation 39
3.3 Overview of Object and Feature Spaces 40
3.3.1 Example: Probability Distributions as Data Objects 43

4 Exploratory and Confirmatory Analyses 47

4.1 Exploratory Analysis–Discover Structure in Data 47
4.1.1 Example: Tilted Parabolas FDA 48
4.1.2 Example: Twin Arches FDA 52
4.1.3 Case Study: Lung Cancer Data 55
4.1.4 Case Study: Pan-Cancer Data 60
4.2 Confirmatory Analysis–Is It Really There? 63
4.3 Further Major Statistical Tasks 69

vii
viii CONTENTS
5 OODA Preprocessing 71
5.1 Visualization of Marginal Distributions 71
5.1.1 Case Study: Spanish Mortality Data 72
5.1.2 Case Study: Drug Discovery Data 74
5.2 Standardization–Appropriate Linear Scaling 85
5.2.1 Example: Two Scale Curve Data 86
5.2.2 Overview of Standardization 89
5.3 Transformation–Appropriate Nonlinear Scaling 91
5.4 Registration–Appropriate Alignment 94

6 Data Visualization 97
6.1 Heat-Map Views of Data Matrices 97
6.2 Curve Views of Matrices and Modes of Variation 104
6.3 Data Centering and Combined Views 107
6.4 Scatterplot Matrix Views of Scores 116
6.5 Alternatives to PCA Directions 120

7 Distance Based Methods 125

7.1 Fréchet Centers In Metric Spaces 127
7.2 Multi-Dimensional Scaling For Object Representation 132
7.3 Important Distance Examples 136
7.3.1 Conventional Norms 136
7.3.2 Wasserstein Distances 137
7.3.3 Procrustes Distances 139
7.3.4 Generalized Procrustes Analysis 141
7.3.5 Covariance Matrix Distances 143

8 Manifold Data Analysis 147

8.1 Directional Data 147
8.2 Introduction to Shape Manifolds 149
8.3 Statistical Analysis of Shapes 151
8.4 Landmark Shapes 157
8.4.1 Shape Tangent Space 160
8.4.2 Case Study: Digit 3 Data 160
8.4.3 Case Study: DNA Molecule Data 162
8.4.4 Principal Nested Shape Spaces 164
8.4.5 Size-and-shape space 166
8.4.6 Further Methodology 167
8.5 Central Limit Theory on Manifolds 167
8.6 Backwards PCA 169
8.7 Covariance Matrices as Data Objects 172

9 FDA Curve Registration 175

9.1 Fisher-Rao Curve Registration 176
9.1.1 Example: Shifted Betas Data 176
CONTENTS ix
9.1.2 Introduction to Warping Functions 181
9.1.3 Fisher-Rao Mathematics 182
9.2 Principal Nested Spheres Decomposition 193

10 Graph Structured Data Objects 197

10.1 Arterial Trees as Data Objects 198
10.1.1 Combinatoric Approaches 198
10.1.2 Phylogenetics 199
10.1.3 Dyck Path 202
10.1.4 Persistent Homology 203
10.1.5 Comparison of Tree Analysis Methods 206
10.2 Networks as Data Objects 207
10.2.1 Graph Laplacians 207
10.2.2 Example: A Tale of Two Cities 209
10.2.3 Extrinsic and Intrinsic Analysis 211
10.2.4 Case Study: Corpus Linguistics 211
10.2.5 Labeled versus Unlabeled Nodes 213

11 Classification–Supervised Learning 215

11.1 Classical Methods 217
11.2 Kernel Methods 226
11.3 Support Vector Machines 232
11.4 Distance Weighted Discrimination 236
11.5 Other Classification Approaches 241

12 Clustering–Unsupervised Learning 243

12.1 K-Means Clustering 243
12.2 Hierarchical Clustering 247
12.3 Visualization Based Methods 254
12.3.1 Hybrid Clustering Methods 256

13 High-Dimensional Inference 257

13.1 DiProPerm–Two Sample Testing 257
13.2 Statistical Significance in Clustering 262
13.2.1 High Dimensional SigClust 266

14 High Dimensional Asymptotics 275

14.1 Random Matrix Theory 276
14.2 High Dimension Low Sample Size 281
14.3 High Dimension Medium Sample Size 290

15 Smoothing and SiZer 293

15.1 Why Not Histograms?–Hidalgo Stamps Data 294
15.2 Smoothing Basics–Bralower Fossils Data 299
15.3 Smoothing Parameter Selection 302
x CONTENTS
15.4 Statistical Inference and SiZer 303
15.4.1 Case Study: British Family Incomes Data 304
15.4.2 Case Study: Bralower Fossils Data 307
15.4.3 Case Study: Mass Flux Data 307
15.4.4 Case Study: Kidney Cancer Data 308
15.4.5 Additional SiZer Applications and Variants 311

16 Robust Methods 313

16.1 Robustness Controversies 314
16.2 Robust Methods for OODA 315
16.2.1 Case Study: Cornea Curvature Data 321
16.2.2 Case Study: Genome-Wide Association Data 325
16.3 Other Robustness Areas 327

17 PCA Details and Variants 331

17.1 Viewpoints of PCA 332
17.1.1 Data Centering 334
17.1.2 Singular Value Decomposition 342
17.1.3 Gaussian Likelihood View 348
17.1.4 PCA Computational Issues 349
17.2 Two Block Decompositions 350
17.2.1 Partial Least Squares 351
17.2.2 Canonical Correlations 354
17.2.3 Joint and Individual Variation Explained 359

18 OODA Context and Related Areas 361

18.1 History and Terminology 361
18.2 OODA Analogy with Object-Oriented Programming 362
18.3 Compositional Data Analysis 364
18.4 Symbolic Data Analysis 365
18.5 Other Research Areas 367

Bibliography 371

Index 416
 Preface

This book is intended as a resource for researchers in the development of novel

statistical and data science methodology. At the time of this writing, Big Data is
a very popular area of study. While Big Data does indeed present major statistical
challenges, an even greater challenge is dealing effectively with Complex Data
which is the main motivation for Object Oriented Data Analysis. The latter is a
framework that facilitates inter-disciplinary research through new terminology for
discussing the often many possible approaches to the analysis of complex data.
Such data are naturally arising in a wide variety of areas. This book aims to pro-
vide ways of thinking that enable the making of sensible choices. The main points
are illustrated with many real data examples, based on the authors’ personal expe-
riences, which have motivated the invention of a wide array of analytic methods.
A generally relevant comment is that most statistical problems can be solved
in many sensible ways. Simon Sheather elegantly summarized that state of affairs
(applicable to the many methods discussed in this book) as: “every dog has its
day”. The point is that any method that has been seriously advocated by someone
has situations where it gives excellent performance, but also situations where it can
be quite poor. The challenge is to understand the properties of each well enough to
guide good choices. The material in this book will provide the reader with useful
insights and a general framework to assist in this process.
A fundamental theme throughout the book, that has not been deeply explored
elsewhere is modes of variation. That provides a novel terminology and frame-
work for understanding many aspects of Object Oriented Data Analysis.
While the mathematics goes far beyond the usual in statistics (including differ-
ential geometry and even topology), the book is aimed at accessibility by graduate
students. There is deliberate focus on ideas over mathematical formulas. An ex-
ception is the detailed linear algebra development in Chapter 17. While many
references to various aspects of OODA are given, it should be noted we have
deliberately not attempted to be comprehensive in those. Our aim instead is to
simply provide useful starting points that interested researchers can use for their
own bibliographic searches.
The historical background of the Object Oriented Data Analysis terminology is
discussed in Section 18.1.
Much of the material that went into this book, including data sets, and the code
to generate most of the graphics can be found in the the web companion to this
book at Marron (2020). Many of those require Marron’s Matlab Software, avail-
able at: Marron (2017b). The companion website also contains further references
and links to other software packages.

xi
xii PREFACE
Acknowledgments
Many of the ideas and presentation style have been developed during the teaching
of a graduate course entitled Object Oriented Data Analysis and have been taught
roughly every other year at the University of North Carolina since 2005. There
were some important precursors, including a related course taught at Cornell Uni-
versity in 2002. Two such courses were offered at the Statistical and Mathemati-
cal Sciences Institute in 2010 and 2011 with lecturers (beyond the authors of this
book) including Hans-Georg Müller, James O. Ramsay, and Jane-Ling Wang. The
course was also taught at the National University of Singapore in 2015.
Several events have played a pivotal role in the development of Object Oriented
Data Analysis. One was the Statistical and Mathematical Sciences Institute pro-
gram on “Analysis of Object Data” during 2010–2011, with co-organizers Hans-
Georg Müller, James O. Ramsay and Jane-Ling Wang. Another pivotal workshop
was the November 2012 “Statistics of Time Warpings and Phase Variations” at the
Mathematical Biosciences Institute, with co-organizers James O. Ramsay, Laura
Sangalli and Anuj Srivastava.
General research in this area by the authors has been supported over the years
by a number of grants from the National Science Foundation, including DMS-
9971649, DMS-0308331, DMS-0606577, DMS-0854908 and IIS-1633074, and
the Engineering and Physical Sciences Research Council grants EP/K022547/1
and EP/T003928/1.
The material on sounds as data objects was kindly provided by Davide Pigoli.
The authors are especially grateful to Stephan F. Huckemann, John T. Kent, James
O. Ramsay, and Anuj Srivastava for providing formal reviews of early drafts that
fundamentally impacted the final version. Additional useful comments on various
drafts have been provided by Iain Carmichael, Benjamin Elztner, Thomas Keefe,
Carson Mosso, Vic Patrangenaru, Stephen M. Pizer, Davide Pigoli, and Piercesare
Secchi.
The authors are grateful to John Kimmel, for helpful advice at many points, and
for his patience over the long time it took for this book to come together.
 CHAPTER 1

What Is OODA?

The fields of human endeavor currently known as statistics, data science, and data
analytics have been radically transformed over the recent past. These transforma-
tions have been driven simultaneously by a massive increase in computational
capabilities coupled with a rapidly growing scientific appetite for ever deeper un-
derstanding and insights. The notion of forming a data matrix provides a useful
paradigm for understanding important aspects of how these fields are evolving.
In particular, the currently popular context of Big Data has several quite different
facets, ranging from low dimension high sample size areas (the basis of classi-
cal mathematical statistical thought, which is perhaps typified by sample survey
and census data), through both high dimension and sample sizes (common for in-
ternet scale data sets of many types), and on to high dimension low sample size
contexts (frequently encountered in areas such as genetics, medical imaging and
other types of extremely rich but relatively expensive measurements). The press-
ing need to analyze data in this wide array of contexts has generated many exciting
new ideas and approaches.
Yet a deeper look into these developments suggests that the organization of data
into a matrix may itself be imposing limitations. In particular, there is a growing
realization that the challenges presented by Big Data are being eclipsed by the
perhaps far greater challenges of Complex Data, which are typically not easily
represented as an unconstrained matrix of numbers. Object Oriented Data Anal-
ysis (OODA) provides a useful general framework for the consideration of many
types of Complex Data. It is deliberately intended to be particularly useful in the
analysis of data in complicated situations, diverse examples of which are given in
the first two chapters. The phrase OODA in this context was coined by Wang and
Marron (2007). An overview of the area was given in Marron and Alonso (2014).
For more discussion of Big Data and its relation to statistics, see Carmichael and
Marron (2018) and many interesting viewpoints in the special issue edited by San-
galli (2018).
The OODA viewpoint is easily understood through taking data objects to be
the atoms of a statistical analysis, where atom is meant in the sense of elementary
particle, studied in several contexts of increasing complexity:
• In a first course in statistics atoms are numbers, and the goal is to develop
methods for understanding of variation in populations of numbers.
• A more advanced course, termed multivariate analysis in the statistical culture,
generalizes the atoms, i.e. the data objects from numbers to vectors and in-
volves a host of methods for managing uncertainty in that context. For example

DOI: 10.1201/9781351189675-1 1
2 WHAT IS OODA?
see Mardia et al. (1979), Muirhead (1982), and Koch (2014) (for a more up to
date treatment).
• At the time of this writing a fashionable area in statistics is Functional Data
Analysis (FDA), where the goal is to analyze the variation in a population of
curves. A good introduction to this vibrant research area, where functions are
the data objects, can be found in Ramsay and Silverman (2002, 2005). A case
study, illustrating many of the basic concepts of FDA which are useful for
understanding OODA is given in Section 1.1.
• OODA provides the next step in terms of complexity of atoms of a statistical
analysis to a wide array of more complicated objects. The important example
of shapes as data objects is considered in the case study of Section 1.2. A wide
variety of other examples, which highlight the breadth of OODA, appears in
Chapter 2.
Note that each of the above areas can be thought of as containing the preced-
ing ones as special cases. For example, multivariate analysis is the case of FDA
where the functions are discretely supported. Similarly multivariate analysis and
FDA are special cases of OODA. In later chapters it is useful to recall that OODA
includes these predecessors as special cases. This is because often simple multi-
variate examples are used for maximal clarity in the illustration of concepts and
methods, but the ideas are useful more generally for OODA.
A good question is: What is the value added to applied statistics and data sci-
ence from the concept of OODA and its attendant terminology? The terminology
is based on very substantial real-world experience with a wide variety of complex
data sets. A fact that rapidly becomes clear in the course of interdisciplinary re-
search is that there frequently are substantial hurdles in terminology. Especially at
the beginning of such endeavors, it can feel like collaborators are even speaking
different languages, so often serious effort needs to be devoted to the develop-
ment of a common set of definitions just to carry on a useful discussion. An added
complication is that for complex data contexts, it is frequently not obvious how to
even “get a handle on the data”. Usually there are many options available, which
are most effectively decided upon through careful discussion between domain sci-
entists and statisticians. In such discussions, the issue of what should be the data
objects? has proven to frequently lead to useful choices, thus resulting in an ef-
fective and insightful data analysis.
Real data examples demonstrating data object choices in a variety of contexts
are given in the following and Chapter 2. In particular, Section 1.1 introduces
curves as data objects. A more complex variant involves curves with interesting
variation in phase in place of, or in addition to, the usual FDA amplitude variation
discussed in Section 2.1. A mathematically deeper case is considered in Section
1.2 where shapes are the data objects which require special treatment as shapes
are most naturally viewed as points on a curved manifold. Section 2.2 considers
a perhaps even more challenging data set of tree-structured data objects. The data
objects in Section 2.3 are recordings of sounds, in particular human spoken words,
Other documents randomly have
different content
 Do not try to make the forward cast on just the same plane as the
back cast for fear that the end of the line should snap like a whip-
lash, which if you were actually fishing would crack off your flies
pretty certainly. Therefore make the lift of the back cast with a slight
sweep (generally inward towards the body is the more natural), and
deliver the forward cast straight out towards its destination. But
always aim about your own height above the spot on the water you
mean to reach to insure the line falling lightly.
In all your practising remember that the key-note of good casting
is in getting a good, clean, high back cast, and in never sending the
line forward until it is quite straight out behind and above you in the
back cast. If you have with you some one to guide you as to when it
is straightened out it will be a great gain, particularly as the time
required for the straightening varies with the length of line that is
used.
Do not try to cast a long line until you have learned to cast a short
one well; and well means not only with a high back cast and straight
forward, but also accurately as to aim and delicately.

Repairs, Knots, and Splices

As has been said already it hardly pays nowadays to make one’s

own tackle, at least for a beginner. But a few things it is useful to
know so that repairs and supplies of a sort can be made in an
emergency. But emergencies are to a great degree prevented by care
—by cleaning and looking over the rod and the reel whenever you
come home, and keeping it safe. The reel may need a drop of oil
now and then, and it should be always kept out of the dust, in a box
or drawer, and above all from falls or blows.
 The repairs most commonly called for are the splicing of a broken
rod, the replacing a lost tip-ring or guide-ring, the knotting of gut,
and the putting a hook to gut. The repair-kit, to use a cyclist’s
phrase, consists of a piece of shoemaker’s wax, some moderately
stout sewing-silk—number A being the best suited for quick work of
all sorts—some bits of flexible brass wire, and a pocket-knife. As the
method of wrapping used in all repairs is essentially the same let us
begin with the simplest, the putting the hook to gut.
First wax well a piece of silk half a yard or more in length. Choose
your hook and the piece of gut (or line if you do not use gut) to
which it is to be fastened. Take the bend of the hook firmly between
the left thumb and forefinger and with the right take two or three
turns of the silk about the other end of the shank (Fig. 4 A). Then lay
the gut on the under side of the hook, reaching two-thirds down the
shank, and wind the silk snugly, coil against coil, over both hook and
gut towards the bend of the hook B. When the gut is nearly covered
make a loop as at C, but relatively larger than in the figure and keep
on winding so that the part a covers in not only hook and gut but the
part b, clearing the silk from the bend of the hook at each turn.
When four or five turns have been taken draw on c and pull it snug.
(D shows the loop not quite drawn down.) When snug cut off the
end and you have the “invisible knot” or “whip-finish” universally
used by fishermen. The same whipping and finish are used for the
other repairs mentioned above.
Suppose that a joint of the rod has broken with a slanting break. It
may be that the surfaces can be fitted together neatly. In this case
they may be joined at once, but if any part has been lost or broken
away, then the broken surfaces must be trimmed and smoothed with
the knife until they do fit. They are spliced thus: Rub the surfaces
with your shoemaker’s wax, press them together, and if you cannot
easily hold them, tie them temporarily with a piece of string, or
perhaps still better, make open coils over the joined parts as at A in
Fig. 4. Then wind back over these coils and the joined parts, making
close, snug coils just as you did on the hook. The whip-finish must
be managed a little differently. Fig. 5 shows how this is done. When
ready to finish drop a loop and make four or five loose turns. Carry
the end under the beginning of the loop, wind down the loose coils
firmly, and pull the end through as before.
A lost tip-ring or guide-ring can be made good by a piece of wire
bent into proper shape and whipped on with well waxed silk thread.
 We need say but a few words about knots. In order to knot your
gut it must be first softened, which is done by laying it in water; it
softens much more rapidly in lukewarm water than in cold. Two
lengths of gut are joined together by lapping the end of one by the
end of the other and making a knot in both together. Fig. 6 A shows
this knot, only the end is put through twice for greater security. If
the fishing is such as does not need strong tackle a single knot will
suffice.
A loop at the end of a piece of gut is made by making the knot as
at B, which is the commonly used knot.
A better knot is that shown at C, which looks complicated as
drawn, but really is not so, as in tying it the two loops a and b are
made, the end c laid between them, and then b is drawn through a.
D shows the ordinary simple method of fastening a line to the loop
of a leader or of a snell. It is the “becket hitch” of the seaman. It
explains itself. Its great advantage is that it cannot slip if drawn
down snugly and can be instantly loosened by pushing the main line
back a little way.
 Aids for Young Anglers

How often has it happened that on reaching a camping-ground,

hotel, or boarding-house near river or lake where pickerel, bass, and
large perch abounded no provision is found for the angler’s sport but
a boat—no lines, sinkers, or floats; no nets for catching live bait, and
no bait but worms. For sunfish, catfish, and small perch, worms are
very fair bait; but for pickerel, bass, and large perch live bait is best.
Here are some makeshifts and aids that may be gotten up at short
notice and at small expense.
Fig. 7 is an end-section of a mosquito-net seine for taking live bait.
The length of the seine is thirty-eight feet, depth five feet. The “cork-
line” A A consists of a small-sized clothes-line. Corks not always
being obtainable, I have used pieces of thoroughly seasoned white
pine three inches in length and one inch in diameter (C C C).
Through these rounded pieces of wood holes are bored through
which the clothes-line passes. These floats are placed eight inches
apart and are kept in position by the clothes-line fitting tightly in the
holes. At the bottom of the seine another clothes-line is sewed to the
netting (B B). This is called the “lead line” and is for the purpose of
keeping the lower part of the seine close to the bottom of the water.
In the lead line pieces of sheet-lead one inch in length are fastened
(H H H) twenty-eight inches apart. The “staff” D is a well-seasoned
piece of hickory six feet long, to the lower end of which sheet-lead is
also fastened at E to keep it down. To the staff is attached the staff
line F F F, thirty feet long, which is for the purpose of drawing in the
seine after it has been cast.
 A seine of this size is generally worked by two persons and two
boats. Each person takes one of the staff lines in his boat, and
rowing towards the shore with the extended seine describes a
semicircle between the boats. As the shore is approached each boat
closes in, thereby causing the two staffs to meet and imprison all the
fish that have come within the bounds of the seine. When one
person works the seine one of the staff lines is tied to a rock or stake
on the shore and the other line is taken into a boat, or the operator
wades out and causes his end of the seine to describe a circle until
the two shafts meet. Great care must be taken to keep the lead line
close to the bottom otherwise the fish will escape. In the selection of
the seining-ground always avoid stony bottoms, snags, and brush,
which will cause the seine to “roll up” and tear.
The cost of the above-described seine ranges from three to four
dollars, and is capable of lasting two seasons if carefully handled and
spread out on the grass to dry after using it. A much superior article
to mosquito-net is bobinet, which will last several seasons.
Fig. 8 is a bait-boat for keeping the bait alive. It is towed behind or
kept by the side when fishing. The top and bottom pieces consist of
half-inch pine; in the centre of each piece square openings are cut;
that on the top is protected by a door made of wire-cloth or quarter-
inch mesh fastened to two small staples which answer the purpose
of hinges; over the opening in the bottom piece wire-cloth is nailed
to admit of a free circulation of water. Under the back end of the top
piece a cleat is nailed, also two cleats on the bottom piece as shown
in the drawing. At the bow of the boat an upright piece of wood is
fastened to the top and bottom of the bait-boat by means of screws.
The sides of the boat consist of one piece of wire-cloth, the ends of
which meet at the upright piece of wood at the bow and are nailed
with broad-headed galvanized nails. The top and bottom of the wire-
cloth are also fastened with nails to the edges of the top and bottom
of the boat as shown in the drawing. A tow-line is fastened to the
bow and the boat is complete.
When handling the bait a small hand-net (Fig. 9) is used,
consisting of a stout piece of wire as shown in the drawing. The
straight parts of the wire are bound together with fishing-line and
constitute the handle; to this frame netting is sewed to form the net-
bag.
For a makeshift float nothing is better than a good-sized bottle-
cork into which a cut has been made with a sharp knife or razor
extending from the side to the centre of the cork. Into this cut the
line is drawn as shown in Fig. 10 A.
Sheet-lead is always a useful aid in makeshift fishing-tackle, and
for light lines makes excellent sinkers when bent and compressed
around the line as shown at Fig. 10 B.
For cleaning out a boat a stiff whisk-broom made of fine birch
twigs bound together with wire or fishing-line, as shown at Fig. 11,
will be found very useful.
Fig. 12 A and B are hand-made sinkers beaten and carved out of
old lead pipe. The carved one, B, is first roughed out with a jack-
knife and finished up with fine emery or sand paper. A is beaten into
shape with a railroad spike on an anvil or smooth stone. This beating
and carving of lead is very pleasant work, the lead being of such an
easy and good-natured temper.
For a cheap and easy-obtainable bailer make use of an empty
tomato or corned-beef can as shown in Fig. 13. A hole sufficiently
large to admit of the handle is punched in the side of the can; the
inside of the handle is chamfered off so as to fit close to the inner
side of the can. Through the can and into the end of the handle a
stout nail is driven as at A.
 SOME USEFUL HINTS

A good bait for large fish is a strip cut from the under side of a
small pickerel, perch, or sunfish, which is placed on the hook as
shown in Fig. 14.
 Baits, and Where to Find Them

As a rule, the young fisherman naturally considers the angle-worm

to be the only bait he need have when he goes fishing, and, taking a
spade, he seeks a moist, loamy spot in the garden and proceeds to
fill his box. But there is a choice even in worms, and those of a clear,
dark, amber color are the best.
Just at night, and after a soft, warm rain, worms of the size of a
lead-pencil will be found crawling over the ground. These are
excellent bait for bass, chub, perch, and large trout early in the
season. If you step very lightly, so as not to jar the ground, you can
easily pick up a box of these large fellows.
It is well enough to have worms with you on all occasions, but
there are many other baits. Frogs, crabs, grubs, dobsons, minnows,
June-bugs, grasshoppers, and crickets, as well as artificial baits, are
more successful lures with certain fish. A few words telling where to
find these baits and the proper manner of keeping them fresh and
lively may prove of value.
Frogs are most plentiful on the shores of ponds or streams filled
with plant growth, and in low, moist places in meadows. In searching
for them in grass, wait till you see one jump, then catch it in your
hands. They are not so easily gotten from the shores, as they are apt
to take to the water at the first alarm.
Crabs are usually found under stones along the shores of a stream
or pond, and in some localities in low, moist places in grass-lands.
Seize the crab back of the pincers and it cannot nip you.
Dobsons are only found under mossy stones in swift-running
waters. They are of a dark-gray color, have many legs, and when
fully grown are about three inches long. The head is shield-shaped
and armed with good, stout pincers, so handle the dobson as you
would a crab. The best way to get a supply of dobsons is to have
some one hold one edge of a fine-meshed net on the bottom of the
stream while you turn over the stones above the net with a hoe. The
dobsons, loosing their hold on the bottom, will be carried by the
current into the net. Put frogs, crabs, and dobsons into a pail with
plenty of grass and some water. If you are to keep them for some
time change the water occasionally.
Grubs are excellent bait for trout early in the season. They are
found in partially decayed tree-trunks, stumps, and old timbers left in
moist places. Cut into the wood with an axe, and if you find it full of
holes of the size of a lead-pencil, knock it to pieces and pick out the
grubs. Put them in a tin bait-box with some of the rotten wood you
found them in.
Minnows of a size suitable for perch and bass fishing can usually
be procured from a spring hole or the pools of a small stream. Take a
rather baggy net with a small mesh, and after setting it at one end of
the pool drive the minnows into it by striking on the water with a
pole and punching about on the bottom. If you stir up the water the
little fish will drive more easily. If your supply must be procured from
a lake or pond, look among the shallows close inshore until you have
found a school, then draw a small seine around them. Large
minnows for pickerel or pike fishing can be caught with a hook and
line. Those you are to use for skittering had better be packed in salt.
The minnows you would keep alive should be put into the bait-pail as
soon as caught. Bait-pails, as usually made, consist of one pail freely
perforated with holes to be set into a tight outer pail. By this
arrangement the water can be changed frequently without
inconveniencing the little fellows. If the bait is to be carried some
distance, and there is no chance to change the water, pack the space
between the two pails loosely with grass. The water trickling down
through the grass will take up the air needed by the fish.
Crickets are to be found under stones, loose sods, and old planks.
Select the largest you can find. June-bugs, sometimes called May-
bugs, hide through the heat of the day among the leaves of the
trees, and sometimes by shaking a tree quite a number will fall to
the ground. Grasshoppers are plentiful in meadow and pasture lands,
and may easily be caught in the hands. Put June-bugs, crickets, and
grasshoppers in a wide-mouthed bottle loosely stuffed with grass. Do
not cork the bottle tight.
 I never esteemed artificial baits, such as the rubber frog and crab,
very highly. It is impossible to give the semblance of life to them in
the water, and most game fish prefer live food to dead. The spoon-
hook and the artificial fly, however, have proved their worth. The
spoon should be of a size in keeping with the size of the game fished
for, and it is well enough to have two—one bright, for use early in
the morning and late in the afternoon and on dark days, the other
dull-colored for use in the brightest part of the day. It is an excellent
plan to bait a spoon-hook with a large worm, a minnow, or a piece of
meat; then if the fish strikes and misses the hook it may get a
portion of the bait and will strike again with truer aim.
There are many other things that can be used for bait, which are
to be found only in your locality. What they are you can learn by
observation and experiment. One can always learn something. Only
recently I discovered that bass were fond of darning-needles.
Sometimes the fish have very fickle appetites, and it is well to have
as many kinds of bait as you can conveniently carry. It is also a good
plan to open the stomach of the first fish you catch, and offer to its
companions the same kind of food found inside of it.

A Trap for Small Fish

Many of the boys and girls who live near the sea-side are
interested in making and stocking aquariums, and many, no doubt,
have experienced the same difficulty which I did when I used to
stock aquariums myself.
I always found that the scoop-net which we use to catch the fish
with is good enough for certain kinds of minnows, but there are
others which are too lively or too shy to be caught in that way; so I
set to work to devise some plan for their capture. I claim no
originality for this trap—it is hundreds of years old; but as it
answered my purpose better than anything else, I used it. The way I
made it was as follows:
 I took a piece of wire-netting about three feet square and bent it
so as to form a tube three feet long and about one foot in diameter
(Fig. 15). I then took two other strips of wire-netting, three feet long
at the top, one foot wide, and two feet at the bottom (Fig. 16); these
I bent into funnel shape. I sewed one funnel in about the middle of
my cylinder and another in one end, as shown in Fig. 15,
strengthening them in their position with strings from the small ends
to the sides of the cylinder. The other end of the cylinder I closed
with a piece of strong bagging so sewed on that there was a space
left at one side which could be untied when I wished to empty the
trap.
The manner of setting the trap is as simple as its manufacture. A
handful of clams or mussels, crushed so that the minnows can get at
the flesh, is thrown in between the first and second funnels. The fish,
little crabs, small eels, and the like, go in, and when they try to get
out they find it much easier to swim through the second funnel than
to find the small hole in the first. I have had several of these traps,
or “pots,” as the fishermen call them, in operation at one time, and
have caught as many as half a bushel of small fish in one night.
The trap can be made by making a frame of hoops and lath and
covering it with mosquito-netting, but it is not so desirable as the
fine wire, being more easily torn.

A Water-turtle Trap

Some time ago, while spending the summer in the country, I

began the pleasing amusement of making an aquarium. I used
various methods to procure the inmates of the great glass box which
I had made for the purpose, and was successful, except that I could
not get a water-turtle. There they would lie on logs in the pond
sunning themselves, but the moment I came within reaching
distance, plump they would go into the water. At last I took an old
soap-box, and after carefully removing one end I nailed on the cover.
I then fastened the end to the cover by hinges, so that it would
swing inward, and after throwing in a few bones and scraps of meat,
I sunk the box in the pond close beside a big log where the turtles
were accustomed to sun themselves. I put a heavy stone on the box,
so as to keep it steady, and awaited the next morning for
developments. Fig. 17.
Here I may say that this trap takes advantage of a peculiarity in
the nature of the water-turtle—namely, if there is a log or stone that
he cannot get under, that is just the place that he wants to get; and I
calculated that the slight resistance offered by my swinging door
would be just enough to make the turtles determined to get into my
box. The next morning when I went to my trap I found several
turtles of all sizes, from one tiny, yellow-spotted fellow, or mud-
turtle, not larger than a half-dollar, to an ugly, great snapper as big
as your hat, and so ill-tempered that I let him go again, glad enough
at having got rid of so troublesome a visitor. After that I set my trap
several times and caught a number of turtles. The smaller ones
furnish a charming addition to an aquarium, and the larger ones, if
properly dressed, make a capital stew.

An Eel-pot

All along the Atlantic coast eel-pots are made on the same general
plan, a bottle-shaped basket having a funnel fitted at the bottom and
provided with a hat that is held on by two straps of green oak.
Three forms are used on which to build up the basket-work. The
large form is usually ten inches in diameter and shaped down to
eight inches at the top or neck. This form is two feet long and has a
round stick driven in the small end. This in turn rests in a hole bored
in a solid piece of plank, so that it is held in an inverted position and
revolves in the hole. Green oak is used for the ribs and bands. This is
cut as straight and free from knots as possible, and is soaked in
water for weeks before it is split and slivered. Green oak will sliver in
an even and uniform manner if it is started right, and from the trunk
of an oak-tree six inches in diameter enough material can be had to
make several dozen eel-pots. The ribs are three-quarters of an inch
wide and about one-eighth of an inch thick, while the bands are a
trifle thinner and wider. A number of the ribs are tied around the
form as shown in Fig. 18, and beginning at the bottom the bands are
woven in and out around the form, turning it as the work progresses
so that the immediate parts are always in sight. Where the ends join
they are shaved down thin so that one laps over the other; then the
weaving continues until the top is reached. The ends of the ribs are
then shaved thin and bent back and slipped under some of the
straps. A thin ribbon of the oak is sewed over and over around the
edge to finish it. The top or small end of the basket is finished in a
similar manner.
The cone or funnel form is fifteen inches long, nine inches in
diameter at the large end, and tapers down to two inches at the
bottom as shown in Fig. 19. Ribs are tied to this form the same as in
the case of the large one, and the weaving begins at the bottom and
is carried to the top, where the ends of the ribs are shaved and
turned in as before described. The bottom or small end of the funnel
is the trap, and here the long, thin ends of the ribs are left, so that
the eel, when he goes through the funnel and into the pot, cannot
get back again.
 The hat is woven the same as a basket by crossing the ribs and
adding a half-rib from the centre anywhere on the circle, so as to
make an uneven number of ribs; thus the weaving will not duplicate
after the first turn around the circle. This extra rib is shown at A in
Fig. 20. A hat form, shown in Fig. 21, is made of wood and mounted
on a block so that it will revolve the same as the other forms. When
a part of the hat is woven it is placed on the form and two small nails
driven through the ribs into the form to hold the weaving in place. It
is then shaped down over the rounded edges of the form and carried
one or two inches below the form so the lower edges of the ribs can
be shaved and bent easily. A long strap of the green oak is passed
under one of the ribs in the hat and caught under bands of the body
as shown in the drawing of a complete eel-pot. Fig. 22.
The funnel is sewed to the bottom edge of the body with thin
bands. As soon as the pots are finished they should be sunk in
shallow water to keep them wet and get them thoroughly water-
soaked.
Stakes or poles are to be driven or worked down into the bottom
of the bay and the eel-pots made fast to them with ropes. To bait an
eel-pot crack some hard-shell crabs or shrimp or put some pieces of
fresh, raw meat within the pot and drop it overboard. Run the pots
morning and night, and remove the eels by unstrapping the cap and
dumping them into a barrel which may be carried on the boat.

A Scap-net

A scap-net for crabbing or landing fish on a hook may be made

from a ring of heavy galvanized iron driven into the end of a hard-
wood stick. Scap-nets may be purchased in most any general store
near a bay or pond, but the ingenious boy can make one himself
from a hoop and a ball of cotton twine. Hang long pieces of string
over the ring and tie them fast with a square knot. Then tie one
string with its next neighbor all around the circle. Begin lower down
and tie them again, and continue in this manner until the net
resembles Fig. 23. When it is seven or eight inches deep begin to
shape it in at the bottom by making the meshes or openings smaller
so that it will have a rounded bottom. The ends of the string should
be tied together or over a small galvanized-iron sail-ring. All the
strings should be tied in square knots so that they will not become
undone after the net has been used for a while.

A Hoop Drop-net
 A hoop drop-net such as shown at Fig. 24 may easily be made
from three galvanized-wire rings and a mesh of tied string as
described for the scap-net. The hoops should be eighteen inches in
diameter and separated ten inches, thus making a net twenty inches
deep. A mesh is to be formed across the bottom, and at the top six
small ropes are tied and the ends brought together fifteen or twenty
inches above the top ring.
Place some crushed crab or any good bait in the bottom of the net
and slowly lower it until the rings rest on the bottom of the bay or
pond, but keep the small ropes clear from the net. Watch through
the water for visitors, and when the right subject is at the bait and
within the rings give a quick jerk and pull the net rapidly to the
surface. If fish are to be caught in this manner the hoops should be
larger and one more added to the net, making it thirty inches deep.
Fish are cunning and swift, and will often dart up and over the top
hoop faster than you can haul it up.
 Chapter X

LAND-YACHTS AND PUSHMOBILES

A Land-yacht

A few years ago the only kind of yacht known to the boys were those
that sailed in the water, but in this advanced time, when many
unheard-of things have been made possible, the land-yacht has
made its welcome appearance. Down on the Southern coasts,
particularly Florida and California, where the sand packs fine and
hard, the land-yacht is an important feature both for pleasure and
business, and if properly handled in a good breeze it will run from
ten to twenty miles an hour. No end of fun can be had with a
properly constructed boat, and the ingenious boy may employ old
baby-carriage or bicycle wheels for the running-gear.
A yacht of medium size can be made on the lines of Fig. 1 at a
comparatively small cost for the timbers and sail-cloth, spars and
hardware. The leg-of-mutton sail is used in preference to the square
sail, as it has the greatest area close to the ground and is less liable
to upset and much easier to handle.
To begin with, obtain some spruce joist clear-grained and free from
knots. They should be two-by-four inches and twelve feet long. Cut
one of them eight feet long and use it for the main cross-piece to
which the front wheels are attached. Form a V of two twelve-foot
joist, and fasten them to the cross-piece as shown in Fig. 2. About
ten inches of each piece should project beyond the cross-piece. The
timbers are bolted fast and at the rear end they are bevelled and
brought together, then bolted through from side to side as shown in
the plan (Fig. 2). Three feet back of the long cross-piece a shorter
timber is set in between the V-shaped frame as shown at A. At the
middle of this timber a hole one inch and a half square is cut and
into it a tenon on the butt-end of the bowsprit fits as shown in Fig. 3.

The bowsprit is seven feet long and is bolted fast to the long
cross-piece. Where the end fits into the timber A two angle-blocks
are nailed fast. Seat-planking is cut and screwed or nailed fast to the
V-shaped frame as shown in both Fig. 1 and Fig. 2. The boards
should be ten inches wide and cut to overhang the timbers an inch
or two at both ends.
If the wheels from an old baby-carriage are to be used the axle
should be cut in half with a hack-saw and each part clamped under
an end of the cross-timber with U-shaped clamps having the ends
threaded and provided with nuts and washers as shown at Fig. 4.
The rear or steering wheel is set in a fork that a blacksmith will make
from strap-iron, and a round piece of the same metal, having a
square-headed upper end, will do for the rudder-post as shown at
Fig. 5. A short axle threaded at both ends and provided with nuts will
hold the wheel in place, and when the post is passed up through a
hole made in the timbers a tiller can be slipped over the square
shoulder and bolted fast so that it will stay in place.
The tiller is of hard-wood two inches broad at the rear end, one
inch in thickness, and tapered so that it will be about an inch square
with the corners rounded where it is grasped by the hand. The
handle part of it should be bound with linen cord to improve the grip.
Give the deck wood-work and timbering a few coats of red, buff, or
light-green paint.
The mast-step is rigged over the forward cross-timber. Two upright
pieces of board twenty inches long and eight inches wide are
attached to the outside edges of the frame-joist with screws. On top
of these a cross-piece is made fast so that the step presents the
appearance of a bench. Two pieces of board six inches wide are
fastened from the corners down to the bowsprit and cross-timber to
brace the step as shown at Fig. 6. An iron brace is made fast to the
top of the step, behind the mast, and to the bowsprit, as may be
seen in the illustration.
 A hole three inches in diameter is cut at the middle of the step-
board, and through this the mast is slipped. A half-inch round iron is
driven into the bottom of the mast after a hole has been bored with
a bit. It should be long enough to project out about two inches. This
steps into a hole bored in the top of the bowsprit, and with the three
wire stays that are caught at the top of the mast and into the eye-
bolts, shown by the arrow-heads in Fig. 2, the mast is held securely
in place.