Experience Seamless Full Ebook Downloads for Every Genre at textbookfull.

com

An Introduction to Generalized Linear Models

Annette J. Dobson

https://textbookfull.com/product/an-introduction-to-
generalized-linear-models-annette-j-dobson/

OR CLICK BUTTON

DOWNLOAD NOW

Explore and download more ebook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Generalized linear models and extensions Fourth Edition

Hardin

https://textbookfull.com/product/generalized-linear-models-and-
extensions-fourth-edition-hardin/

textboxfull.com

Linear and Generalized Linear Mixed Models and Their

Applications 2nd Edition Jiming Jiang

https://textbookfull.com/product/linear-and-generalized-linear-mixed-
models-and-their-applications-2nd-edition-jiming-jiang/

textboxfull.com

Generalized Linear Models and Extensions 4th Edition James

W. Hardin

https://textbookfull.com/product/generalized-linear-models-and-
extensions-4th-edition-james-w-hardin/

textboxfull.com

Generalized Additive Models: An Introduction With R

(Second Edition) Simon N. Wood

https://textbookfull.com/product/generalized-additive-models-an-
introduction-with-r-second-edition-simon-n-wood/

textboxfull.com
Data analysis using hierarchical generalized linear models
with R 1st Edition Youngjo Lee

https://textbookfull.com/product/data-analysis-using-hierarchical-
generalized-linear-models-with-r-1st-edition-youngjo-lee/

textboxfull.com

Repeated Measures Design with Generalized Linear Mixed

Models for Randomized Controlled Trials 1st Edition
Toshiro Tango
https://textbookfull.com/product/repeated-measures-design-with-
generalized-linear-mixed-models-for-randomized-controlled-trials-1st-
edition-toshiro-tango/
textboxfull.com

An Introduction to Orthodontics Simon J. Littlewood

https://textbookfull.com/product/an-introduction-to-orthodontics-
simon-j-littlewood/

textboxfull.com

Numerical Linear Algebra An Introduction 1st Edition

Holger Wendland

https://textbookfull.com/product/numerical-linear-algebra-an-
introduction-1st-edition-holger-wendland/

textboxfull.com

Earth: An Introduction to Physical Geology, 13e Edward J.

Tarbuck

https://textbookfull.com/product/earth-an-introduction-to-physical-
geology-13e-edward-j-tarbuck/

textboxfull.com
An Introduction to
Generalized Linear
Models
Fourth Edition
CHAPMAN & HALL/CRC
Texts in Statistical Science Series
Series Editors
Joseph K. Blitzstein, Harvard University, USA
Julian J. Faraway, University of Bath, UK
Martin Tanner, Northwestern University, USA
Jim Zidek, University of British Columbia, Canada
Nonlinear Time Series: Theory, Methods, Statistical Rethinking: A Bayesian Course
and Applications with R Examples with Examples in R and Stan
R. Douc, E. Moulines, and D.S. Stoffer R. McElreath
Stochastic Modeling and Mathematical Analysis of Variance, Design, and
Statistics: A Text for Statisticians and Regression: Linear Modeling for Unbalanced
Quantitative Scientists Data, Second Edition
F.J. Samaniego R. Christensen
Introduction to Multivariate Analysis: Essentials of Probability Theory for
Linear and Nonlinear Modeling Statisticians
S. Konishi M.A. Proschan and P.A. Shaw
Linear Algebra and Matrix Analysis for Extending the Linear Model with R:
Statistics Generalized Linear, Mixed Effects and
S. Banerjee and A. Roy Nonparametric Regression Models, Second
Bayesian Networks: With Examples in R Edition
M. Scutari and J.-B. Denis J.J. Faraway
Linear Models with R, Second Edition Modeling and Analysis of Stochastic Systems,
J.J. Faraway Third Edition
V.G. Kulkarni
Introduction to Probability
J. K. Blitzstein and J. Hwang Pragmatics of Uncertainty
J.B. Kadane
Analysis of Categorical Data with R
C. R. Bilder and T. M. Loughin Stochastic Processes: From Applications to
Theory
Statistical Inference: An Integrated P.D Moral and S. Penev
Approach, Second Edition
H. S. Migon, D. Gamerman, and F. Louzada Modern Data Science with R
B.S. Baumer, D.T Kaplan, and N.J. Horton
Modelling Survival Data in Medical
Research, Third Edition Logistic Regression Models
D. Collett J.M. Hilbe
Design and Analysis of Experiments with R Generalized Additive Models: An
J. Lawson Introduction with R, Second Edition
S. Wood
Mathematical Statistics: Basic Ideas and
Selected Topics, Volume I, Second Edition Design of Experiments: An Introduction
P. J. Bickel and K. A. Doksum Based on Linear Models
Max Morris
Statistics for Finance
E. Lindström, H. Madsen, and J. N. Nielsen Introduction to Statistical Methods for
Financial Models
Spatio-Temporal Methods in Environmental T. A. Severini
Epidemiology
G. Shaddick and J.V. Zidek Statistical Regression and Classification:
From Linear Models to Machine Learning
Mathematical Statistics: Basic Ideas and N. Matloff
Selected Topics, Volume II
P. J. Bickel and K. A. Doksum Introduction to Functional Data Analysis
P. Kokoszka and M. Reimherr
Mathematical Statistics: Basic Ideas and
Selected Topics, Volume II Stochastic Processes: An Introduction, Third
P. J. Bickel and K. A. Doksum Edition
P.W. Jones and P. Smith
Discrete Data Analysis with R: Visualization
and Modeling Techniques for Categorical
and Count Data
M. Friendly and D. Meyer
An Introduction to
Generalized Linear
Models
Fourth Edition

By
Annette J. Dobson
and
Adrian G. Barnett
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2018 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed on acid-free paper
Version Date: 20180306
International Standard Book Number-13: 978-1-138-74168-3 (Hardback)
International Standard Book Number-13: 978-1-138-74151-5 (Paperback)
This book contains information obtained from authentic and highly regarded sources. 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, please access
www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc.
(CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization
that provides licenses and registration for a variety of users. For organizations that have been granted
a photocopy license by the CCC, a separate system of payment has been arranged.

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: Dobson, Annette J., 1945- author. | Barnett, Adrian G., author.
Title: An introduction to generalized linear models / by Annette J. Dobson,
Adrian G. Barnett.
Other titles: Generalized linear models
Description: Fourth edition. | Boca Raton : CRC Press, 2018. | Includes
bibliographical references and index.
Identifiers: LCCN 2018002845| ISBN 9781138741683 (hardback : alk. paper) |
ISBN 9781138741515 (pbk. : alk. paper) | ISBN 9781315182780 (e-book : alk.
paper)
Subjects: LCSH: Linear models (Statistics)
Classification: LCC QA276 .D589 2018 | DDC 519.5--dc23
LC record available at https://lccn.loc.gov/2018002845

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To Beth.
 Contents

Preface xv

1 Introduction 1
1.1 Background 1
1.2 Scope 1
1.3 Notation 6
1.4 Distributions related to the Normal distribution 8
1.4.1 Normal distributions 8
1.4.2 Chi-squared distribution 9
1.4.3 t-distribution 10
1.4.4 F-distribution 10
1.4.5 Some relationships between distributions 11
1.5 Quadratic forms 11
1.6 Estimation 13
1.6.1 Maximum likelihood estimation 13
1.6.2 Example: Poisson distribution 15
1.6.3 Least squares estimation 15
1.6.4 Comments on estimation 16
1.6.5 Example: Tropical cyclones 17
1.7 Exercises 17

2 Model Fitting 21
2.1 Introduction 21
2.2 Examples 21
2.2.1 Chronic medical conditions 21
2.2.2 Example: Birthweight and gestational age 25
2.3 Some principles of statistical modelling 35
2.3.1 Exploratory data analysis 35
2.3.2 Model formulation 36
2.3.3 Parameter estimation 36

vii
viii
2.3.4 Residuals and model checking 36
2.3.5 Inference and interpretation 39
2.3.6 Further reading 40
2.4 Notation and coding for explanatory variables 40
2.4.1 Example: Means for two groups 41
2.4.2 Example: Simple linear regression for two groups 42
2.4.3 Example: Alternative formulations for comparing the
means of two groups 42
2.4.4 Example: Ordinal explanatory variables 43
2.5 Exercises 44

3 Exponential Family and Generalized Linear Models 49

3.1 Introduction 49
3.2 Exponential family of distributions 50
3.2.1 Poisson distribution 51
3.2.2 Normal distribution 52
3.2.3 Binomial distribution 52
3.3 Properties of distributions in the exponential family 53
3.4 Generalized linear models 56
3.5 Examples 58
3.5.1 Normal linear model 58
3.5.2 Historical linguistics 58
3.5.3 Mortality rates 59
3.6 Exercises 61

4 Estimation 65
4.1 Introduction 65
4.2 Example: Failure times for pressure vessels 65
4.3 Maximum likelihood estimation 70
4.4 Poisson regression example 73
4.5 Exercises 76

5 Inference 79
5.1 Introduction 79
5.2 Sampling distribution for score statistics 81
5.2.1 Example: Score statistic for the Normal distribution 82
5.2.2 Example: Score statistic for the Binomial distribution 82
5.3 Taylor series approximations 83
5.4 Sampling distribution for maximum likelihood estimators 84
 ix
5.4.1 Example: Maximum likelihood estimators for the
Normal linear model 85
5.5 Log-likelihood ratio statistic 86
5.6 Sampling distribution for the deviance 87
5.6.1 Example: Deviance for a Binomial model 88
5.6.2 Example: Deviance for a Normal linear model 89
5.6.3 Example: Deviance for a Poisson model 91
5.7 Hypothesis testing 92
5.7.1 Example: Hypothesis testing for a Normal linear
model 94
5.8 Exercises 95

6 Normal Linear Models 97

6.1 Introduction 97
6.2 Basic results 98
6.2.1 Maximum likelihood estimation 98
6.2.2 Least squares estimation 98
6.2.3 Deviance 99
6.2.4 Hypothesis testing 99
6.2.5 Orthogonality 100
6.2.6 Residuals 101
6.2.7 Other diagnostics 102
6.3 Multiple linear regression 104
6.3.1 Example: Carbohydrate diet 104
6.3.2 Coefficient of determination, R2 108
6.3.3 Model selection 111
6.3.4 Collinearity 118
6.4 Analysis of variance 119
6.4.1 One-factor analysis of variance 119
6.4.2 Two-factor analysis of variance 126
6.5 Analysis of covariance 132
6.6 General linear models 135
6.7 Non-linear associations 137
6.7.1 PLOS Medicine journal data 138
6.8 Fractional polynomials 141
6.9 Exercises 143
x
7 Binary Variables and Logistic Regression 149
7.1 Probability distributions 149
7.2 Generalized linear models 150
7.3 Dose response models 151
7.3.1 Example: Beetle mortality 154
7.4 General logistic regression model 158
7.4.1 Example: Embryogenic anthers 159
7.5 Goodness of fit statistics 162
7.6 Residuals 166
7.7 Other diagnostics 167
7.8 Example: Senility and WAIS 168
7.9 Odds ratios and prevalence ratios 171
7.10 Exercises 174

8 Nominal and Ordinal Logistic Regression 179

8.1 Introduction 179
8.2 Multinomial distribution 180
8.3 Nominal logistic regression 181
8.3.1 Example: Car preferences 183
8.4 Ordinal logistic regression 188
8.4.1 Cumulative logit model 189
8.4.2 Proportional odds model 189
8.4.3 Adjacent categories logit model 190
8.4.4 Continuation ratio logit model 191
8.4.5 Comments 192
8.4.6 Example: Car preferences 192
8.5 General comments 193
8.6 Exercises 194

9 Poisson Regression and Log-Linear Models 197

9.1 Introduction 197
9.2 Poisson regression 198
9.2.1 Example of Poisson regression: British doctors’
smoking and coronary death 201
9.3 Examples of contingency tables 204
9.3.1 Example: Cross-sectional study of malignant
melanoma 205
9.3.2 Example: Randomized controlled trial of influenza
vaccine 206
 xi
9.3.3 Example: Case–control study of gastric and duodenal
ulcers and aspirin use 207
9.4 Probability models for contingency tables 209
9.4.1 Poisson model 209
9.4.2 Multinomial model 209
9.4.3 Product multinomial models 210
9.5 Log-linear models 210
9.6 Inference for log-linear models 212
9.7 Numerical examples 212
9.7.1 Cross-sectional study of malignant melanoma 212
9.7.2 Case–control study of gastric and duodenal ulcer and
aspirin use 215
9.8 Remarks 216
9.9 Exercises 217

10 Survival Analysis 223

10.1 Introduction 223
10.2 Survivor functions and hazard functions 225
10.2.1 Exponential distribution 226
10.2.2 Proportional hazards models 227
10.2.3 Weibull distribution 228
10.3 Empirical survivor function 230
10.3.1 Example: Remission times 231
10.4 Estimation 233
10.4.1 Example: Exponential model 234
10.4.2 Example: Weibull model 235
10.5 Inference 236
10.6 Model checking 236
10.7 Example: Remission times 238
10.8 Exercises 240

11 Clustered and Longitudinal Data 245

11.1 Introduction 245
11.2 Example: Recovery from stroke 247
11.3 Repeated measures models for Normal data 253
11.4 Repeated measures models for non-Normal data 257
11.5 Multilevel models 259
11.6 Stroke example continued 262
11.7 Comments 265
11.8 Exercises 266
xii
12 Bayesian Analysis 271
12.1 Frequentist and Bayesian paradigms 271
12.1.1 Alternative definitions of p-values and confidence
intervals 271
12.1.2 Bayes’ equation 272
12.1.3 Parameter space 273
12.1.4 Example: Schistosoma japonicum 273
12.2 Priors 275
12.2.1 Informative priors 276
12.2.2 Example: Sceptical prior 276
12.2.3 Example: Overdoses amongst released prisoners 279
12.3 Distributions and hierarchies in Bayesian analysis 281
12.4 WinBUGS software for Bayesian analysis 281
12.5 Exercises 284

13 Markov Chain Monte Carlo Methods 287

13.1 Why standard inference fails 287
13.2 Monte Carlo integration 287
13.3 Markov chains 289
13.3.1 The Metropolis–Hastings sampler 291
13.3.2 The Gibbs sampler 293
13.3.3 Comparing a Markov chain to classical maximum
likelihood estimation 295
13.3.4 Importance of parameterization 299
13.4 Bayesian inference 300
13.5 Diagnostics of chain convergence 302
13.5.1 Chain history 302
13.5.2 Chain autocorrelation 304
13.5.3 Multiple chains 305
13.6 Bayesian model fit: the deviance information criterion 306
13.7 Exercises 308

14 Example Bayesian Analyses 315

14.1 Introduction 315
14.2 Binary variables and logistic regression 316
14.2.1 Prevalence ratios for logistic regression 319
14.3 Nominal logistic regression 322
14.4 Latent variable model 324
14.5 Survival analysis 326
14.6 Random effects 328
 xiii
14.7 Longitudinal data analysis 331
14.8 Bayesian model averaging 338
14.8.1 Example: Stroke recovery 340
14.8.2 Example: PLOS Medicine journal data 340
14.9 Some practical tips for WinBUGS 342
14.10 Exercises 344

Postface 347

Appendix 355

Software 357

References 359

Index 371
 Preface

The original purpose of the book was to present a unified theoretical and
conceptual framework for statistical modelling in a way that was accessible
to undergraduate students and researchers in other fields.
The second edition was expanded to include nominal and ordinal logistic
regression, survival analysis and analysis of longitudinal and clustered data.
It relied more on numerical methods, visualizing numerical optimization and
graphical methods for exploratory data analysis and checking model fit.
The third edition added three chapters on Bayesian analysis for general-
ized linear models. To help with the practical application of generalized linear
models, Stata, R and WinBUGS code were added.
This fourth edition includes new sections on the common problems of
model selection and non-linear associations. Non-linear associations have a
long history in statistics as the first application of the least squares method
was when Gauss correctly predicted the non-linear orbit of an asteroid in
1801.
Statistical methods are essential for many fields of research, but a
widespread lack of knowledge of their correct application is creating inaccu-
rate results. Untrustworthy results undermine the scientific process of using
data to make inferences and inform decisions. There are established practices
for creating reproducible results which are covered in a new Postface to this
edition.
The data sets and outline solutions of the exercises are available on
the publisher’s website: http://www.crcpress.com/9781138741515. We also
thank Thomas Haslwanter for providing a set of solutions using Python:
https://github.com/thomas-haslwanter/dobson.
We are grateful to colleagues and students at the Universities of Queens-
land and Newcastle, Australia, and those taking postgraduate courses through
the Biostatistics Collaboration of Australia for their helpful suggestions and
comments about the material.
Annette J. Dobson and Adrian G. Barnett
Brisbane, Australia

xv
 Chapter 1

Introduction

1.1 Background
This book is designed to introduce the reader to generalized linear models,
these provide a unifying framework for many commonly used statistical tech-
niques. They also illustrate the ideas of statistical modelling.
The reader is assumed to have some familiarity with classical statistical
principles and methods. In particular, understanding the concepts of estima-
tion, sampling distributions and hypothesis testing is necessary. Experience
in the use of t-tests, analysis of variance, simple linear regression and chi-
squared tests of independence for two-dimensional contingency tables is as-
sumed. In addition, some knowledge of matrix algebra and calculus is re-
quired.
The reader will find it necessary to have access to statistical computing
facilities. Many statistical programs, languages or packages can now perform
the analyses discussed in this book. Often, however, they do so with a dif-
ferent program or procedure for each type of analysis so that the unifying
structure is not apparent.
Some programs or languages which have procedures consistent with the
approach used in this book are Stata, R, S-PLUS, SAS and Genstat. For
Chapters 13 to 14, programs to conduct Markov chain Monte Carlo methods
are needed and WinBUGS has been used here. This list is not comprehensive
as appropriate modules are continually being added to other programs.
In addition, anyone working through this book may find it helpful to be
able to use mathematical software that can perform matrix algebra, differen-
tiation and iterative calculations.

1.2 Scope
The statistical methods considered in this book all involve the analysis of
relationships between measurements made on groups of subjects or objects.

1
2 INTRODUCTION
For example, the measurements might be the heights or weights and the ages
of boys and girls, or the yield of plants under various growing conditions.
We use the terms response, outcome or dependent variable for measure-
ments that are free to vary in response to other variables called explanatory
variables or predictor variables or independent variables—although this
last term can sometimes be misleading. Responses are regarded as random
variables. Explanatory variables are usually treated as though they are non-
random measurements or observations; for example, they may be fixed by the
experimental design.
Responses and explanatory variables are measured on one of the follow-
ing scales.
1. Nominal classifications: e.g., red, green, blue; yes, no, do not know, not
applicable. In particular, for binary, dichotomous or binomial variables
there are only two categories: male, female; dead, alive; smooth leaves,
serrated leaves. If there are more than two categories the variable is called
polychotomous, polytomous or multinomial.
2. Ordinal classifications in which there is some natural order or ranking be-
tween the categories: e.g., young, middle aged, old; diastolic blood pres-
sures grouped as ≤ 70, 71–90, 91–110, 111–130, ≥ 131 mmHg.
3. Continuous measurements where observations may, at least in theory, fall
anywhere on a continuum: e.g., weight, length or time. This scale includes
both interval scale and ratio scale measurements—the latter have a well-
defined zero. A particular example of a continuous measurement is the time
until a specific event occurs, such as the failure of an electronic component;
the length of time from a known starting point is called the failure time.
Nominal and ordinal data are sometimes called categorical or discrete
variables and the numbers of observations, counts or frequencies in each
category are usually recorded. For continuous data the individual measure-
ments are recorded. The term quantitative is often used for a variable mea-
sured on a continuous scale and the term qualitative for nominal and some-
times for ordinal measurements. A qualitative, explanatory variable is called
a factor and its categories are called the levels for the factor. A quantitative
explanatory variable is sometimes called a covariate.
Methods of statistical analysis depend on the measurement scales of the
response and explanatory variables.
This book is mainly concerned with those statistical methods which are
relevant when there is just one response variable although there will usu-
ally be several explanatory variables. The responses measured on different
subjects are usually assumed to be statistically independent random variables
SCOPE 3
although this requirement is dropped in Chapter 11, which is about correlated
data, and in subsequent chapters. Table 1.1 shows the main methods of statis-
tical analysis for various combinations of response and explanatory variables
and the chapters in which these are described. The last three chapters are de-
voted to Bayesian methods which substantially extend these analyses.
The present chapter summarizes some of the statistical theory used
throughout the book. Chapters 2 through 5 cover the theoretical framework
that is common to the subsequent chapters. Later chapters focus on methods
for analyzing particular kinds of data.
Chapter 2 develops the main ideas of classical or frequentist statistical
modelling. The modelling process involves four steps:
1. Specifying models in two parts: equations linking the response and ex-
planatory variables, and the probability distribution of the response vari-
able.
2. Estimating fixed but unknown parameters used in the models.
3. Checking how well the models fit the actual data.
4. Making inferences; for example, calculating confidence intervals and test-
ing hypotheses about the parameters.
The next three chapters provide the theoretical background. Chapter 3 is
about the exponential family of distributions, which includes the Normal,
Poisson and Binomial distributions. It also covers generalized linear models
(as defined by Nelder and Wedderburn (1972)). Linear regression and many
other models are special cases of generalized linear models. In Chapter 4
methods of classical estimation and model fitting are described.
Chapter 5 outlines frequentist methods of statistical inference for gener-
alized linear models. Most of these methods are based on how well a model
describes the set of data. For example, hypothesis testing is carried out by
first specifying alternative models (one corresponding to the null hypothesis
and the other to a more general hypothesis). Then test statistics are calculated
which measure the “goodness of fit” of each model and these are compared.
Typically the model corresponding to the null hypothesis is simpler, so if it
fits the data about as well as a more complex model it is usually preferred on
the grounds of parsimony (i.e., we retain the null hypothesis).
Chapter 6 is about multiple linear regression and analysis of variance
(ANOVA). Regression is the standard method for relating a continuous re-
sponse variable to several continuous explanatory (or predictor) variables.
ANOVA is used for a continuous response variable and categorical or qual-
itative explanatory variables (factors). Analysis of covariance (ANCOVA)
is used when at least one of the explanatory variables is continuous. Nowa-
4 INTRODUCTION

Table 1.1 Major methods of statistical analysis for response and explanatory vari-
ables measured on various scales and chapter references for this book. Extensions of
these methods from a Bayesian perspective are illustrated in Chapters 12–14.
Response (chapter) Explanatory variables Methods
Continuous Binary t-test
(Chapter 6)
Nominal, >2 categories Analysis of variance

Ordinal Analysis of variance

Continuous Multiple regression

Nominal & some Analysis of

continuous covariance

Categorical & continuous Multiple regression

Binary Categorical Contingency tables
(Chapter 7) Logistic regression

Continuous Logistic, probit &

other dose-response
models

Categorical & continuous Logistic regression

Nominal with Nominal Contingency tables
>2 categories
(Chapters 8 & 9) Categorical & continuous Nominal logistic
regression
Ordinal Categorical & continuous Ordinal logistic
(Chapter 8) regression
Counts Categorical Log-linear models
(Chapter 9)
Categorical & continuous Poisson regression
Failure times Categorical & continuous Survival analysis
(Chapter 10) (parametric)
Correlated Categorical & continuous Generalized
responses estimating equations
(Chapter 11) Multilevel models
SCOPE 5
days it is common to use the same computational tools for all such situations.
The terms multiple regression or general linear model are used to cover the
range of methods for analyzing one continuous response variable and mul-
tiple explanatory variables. This chapter also includes a section on model
selection that is also applicable for other types of generalized linear models
Chapter 7 is about methods for analyzing binary response data. The most
common one is logistic regression which is used to model associations be-
tween the response variable and several explanatory variables which may be
categorical or continuous. Methods for relating the response to a single con-
tinuous variable, the dose, are also considered; these include probit anal-
ysis which was originally developed for analyzing dose-response data from
bioassays. Logistic regression has been generalized to include responses with
more than two nominal categories (nominal, multinomial, polytomous or
polychotomous logistic regression) or ordinal categories (ordinal logistic
regression). These methods are discussed in Chapter 8.
Chapter 9 concerns count data. The counts may be frequencies displayed
in a contingency table or numbers of events, such as traffic accidents, which
need to be analyzed in relation to some “exposure” variable such as the num-
ber of motor vehicles registered or the distances travelled by the drivers. Mod-
elling methods are based on assuming that the distribution of counts can be
described by the Poisson distribution, at least approximately. These methods
include Poisson regression and log-linear models.
Survival analysis is the usual term for methods of analyzing failure time
data. The parametric methods described in Chapter 10 fit into the framework
of generalized linear models although the probability distribution assumed
for the failure times may not belong to the exponential family.
Generalized linear models have been extended to situations where the re-
sponses are correlated rather than independent random variables. This may
occur, for instance, if they are repeated measurements on the same sub-
ject or measurements on a group of related subjects obtained, for example,
from clustered sampling. The method of generalized estimating equations
(GEEs) has been developed for analyzing such data using techniques analo-
gous to those for generalized linear models. This method is outlined in Chap-
ter 11 together with a different approach to correlated data, namely multilevel
modelling in which some parameters are treated as random variables rather
than fixed but unknown constants. Multilevel modelling involves both fixed
and random effects (mixed models) and relates more closely to the Bayesian
approach to statistical analysis.
The main concepts and methods of Bayesian analysis are introduced in
Chapter 12. In this chapter the relationships between classical or frequentist
6 INTRODUCTION
methods and Bayesian methods are outlined. In addition the software Win-
BUGS which is used to fit Bayesian models is introduced.
Bayesian models are usually fitted using computer-intensive methods
based on Markov chains simulated using techniques based on random num-
bers. These methods are described in Chapter 13. This chapter uses some
examples from earlier chapters to illustrate the mechanics of Markov chain
Monte Carlo (MCMC) calculations and to demonstrate how the results allow
much richer statistical inferences than are possible using classical methods.
Chapter 14 comprises several examples, introduced in earlier chapters,
which are reworked using Bayesian analysis. These examples are used to il-
lustrate both conceptual issues and practical approaches to estimation, model
fitting and model comparisons using WinBUGS.
Finally there is a Postscript that summarizes the principles of good
statistical practice that should always be used in order to address the
“reproducibility crisis” that plagues science with daily reports of “break-
throughs” that turn out to be useless or untrue.
Further examples of generalized linear models are discussed in the books
by McCullagh and Nelder (1989), Aitkin et al. (2005) and Myers et al. (2010).
Also there are many books about specific generalized linear models such as
Agresti (2007, 2013), Collett (2003, 2014), Diggle et al. (2002), Goldstein
(2011), Hilbe (2015) and Hosmer et al. (2013).

1.3 Notation
Generally we follow the convention of denoting random variables by upper-
case italic letters and observed values by the corresponding lowercase letters.
For example, the observations y1 , y2 , ..., yn are regarded as realizations of the
random variables Y1 ,Y2 , . . . ,Yn . Greek letters are used to denote parameters
and the corresponding lowercase Roman letters are used to denote estimators
and estimates; occasionally the symbol b is used for estimators or estimates.
For example, the parameter β is estimated by βb or b. Sometimes these con-
ventions are not strictly adhered to, either to avoid excessive notation in cases
where the meaning should be apparent from the context, or when there is a
strong tradition of alternative notation (e.g., e or ε for random error terms).
Vectors and matrices, whether random or not, are denoted by boldface
lower- and uppercase letters, respectively. Thus, y represents a vector of ob-
servations
NOTATION 7
 
y1
 .. 
 . 
yn
or a vector of random variables
 
Y1
 .. 
 . ,
Yn

β denotes a vector of parameters and X is a matrix. The superscript T is

used for a matrix transpose or when a column vector is written as a row, e.g.,
y = [Y1 , . . . ,Yn ]T .
The probability density function of a continuous random variable Y (or
the probability mass function if Y is discrete) is referred to simply as a prob-
ability distribution and denoted by

f (y; θ )

where θ represents the parameters of the distribution.

We use dot (·) subscripts for summation and bars (− ) for means; thus,

1 N 1
y= ∑
N i=1
yi = y · .
N

The expected value and variance of a random variable Y are denoted by

E(Y ) and var(Y ), respectively. Suppose random variables Y1 , . . . ,YN are in-
dependent with E(Yi ) = µi and var(Yi ) = σi2 for i = 1, . . . , n. Let the random
variable W be a linear combination of the Yi ’s

W = a1Y1 + a2Y2 + . . . + anYn , (1.1)

where the ai ’s are constants. Then the expected value of W is

E(W ) = a1 µ1 + a2 µ2 + . . . + an µn (1.2)

and its variance is

var(W ) = a21 σ12 + a22 σ22 + . . . + a2n σn2 . (1.3)

8 INTRODUCTION
1.4 Distributions related to the Normal distribution
The sampling distributions of many of the estimators and test statistics used
in this book depend on the Normal distribution. They do so either directly be-
cause they are derived from Normally distributed random variables or asymp-
totically, via the Central Limit Theorem for large samples. In this section we
give definitions and notation for these distributions and summarize the re-
lationships between them. The exercises at the end of the chapter provide
practice in using these results which are employed extensively in subsequent
chapters.

1.4.1 Normal distributions

1. If the random variable Y has the Normal distribution with mean µ and
variance σ 2 , its probability density function is
"  2 #
1 1 y − µ
f (y; µ , σ 2 ) = √ exp − .
2πσ 2 2 σ
We denote this by Y ∼ N(µ , σ 2 ).
2. The Normal distribution with µ = 0 and σ 2 = 1, Y ∼ N(0, 1), is called the
standard Normal distribution.
3. Let Y1 , . . . ,Yn denote Normally distributed random variables with Yi ∼
N(µi , σi2 ) for i = 1, . . . , n and let the covariance of Yi and Y j be denoted
by
cov(Yi ,Y j ) = ρi j σi σ j ,
where ρi j is the correlation coefficient for Yi and Y j . Then the joint dis-
tribution of the Yi ’s is the multivariate Normal distribution with mean
vector µ = [µ1 , . . . , µn ]T and variance-covariance matrix V with diagonal
elements σi2 and non-diagonal elements ρi j σi σ j for i 6= j. We write this as
y ∼ MVN(µ , V), where y = [Y1 , . . . ,Yn ]T .
4. Suppose the random variables Y1 , . . . ,Yn are independent and Normally dis-
tributed with the distributions Yi ∼ N(µi , σi2 ) for i = 1, . . . , n. If
W = a1Y1 + a2Y2 + . . . + anYn ,
where the ai ’s are constants, then W is also Normally distributed, so that
!
n n n
W = ∑ aiYi ∼ N ∑ ai µi , ∑ a2i σi2
i=1 i=1 i=1

by Equations (1.2) and (1.3).

DISTRIBUTIONS RELATED TO THE NORMAL DISTRIBUTION 9
1.4.2 Chi-squared distribution
1. The central chi-squared distribution with n degrees of freedom is defined
as the sum of squares of n independent random variables Z1 , . . . , Zn each
with the standard Normal distribution. It is denoted by
n
X 2 = ∑ Zi2 ∼ χ 2 (n).
i=1

In matrix notation, if z = [Z1 , . . . , Zn ]T , then zT z = ∑ni=1 Zi2 so that X 2 =

zT z ∼ χ 2 (n).
2. If X 2 has the distribution χ 2 (n), then its expected value is E(X 2 ) = n and
its variance is var(X 2 ) = 2n.
3. If Y1 , . . . ,Yn are independent, Normally distributed random variables, each
with the distribution Yi ∼ N(µi , σi2 ), then
n  2
2 Yi − µi
X =∑ ∼ χ 2 (n) (1.4)
i=1 σi

because each of the variables Zi = (Yi − µi ) /σi has the standard Normal
distribution N(0, 1).
4. Let Z1 , . . . , Zn be independent random variables each with the distribution
N(0, 1) and let Yi = Zi + µi , where at least one of the µi ’s is non-zero. Then
the distribution of

∑ Yi2 = ∑ (Zi + µi )2 = ∑ Zi2 + 2 ∑ Zi µi + ∑ µi2

has larger mean n + λ and larger variance 2n + 4λ than χ 2 (n) where
λ = ∑ µi2 . This is called the non-central chi-squared distribution with
n degrees of freedom and non-centrality parameter λ . It is denoted by
χ 2 (n, λ ).
5. Suppose that the Yi ’s are not necessarily independent and the vector
y = [Y1 , . . . ,Yn ]T has the multivariate Normal distribution y ∼ MVN(µ , V)
where the variance–covariance matrix V is non-singular and its inverse is
V−1 . Then
X 2 = (y − µ )T V−1 (y − µ ) ∼ χ 2 (n). (1.5)
6. More generally if y ∼ MVN( µ , V), then the random variable yT V−1 y has
the non-central chi-squared distribution χ 2 (n, λ ) where λ = µ T V−1 µ .
7. If X12 , . . . , Xm2 are m independent random variables with the chi-squared dis-
tributions Xi2 ∼ χ 2 (ni , λi ), which may or may not be central, then their sum
10 INTRODUCTION
also has a chi-squared distribution with ∑ ni degrees of freedom and non-
centrality parameter ∑ λi , that is,
!
m m m
∑ Xi2 ∼ χ 2 ∑ ni , ∑ λi .
i=1 i=1 i=1

This is called the reproductive property of the chi-squared distribution.

8. Let y ∼ MVN(µ , V), where y has n elements but the Yi ’s are not indepen-
dent so that the number k of linearly independent rows (or columns) of V
(that is, the rank of V) is less than n and so V is singular and its inverse
is not uniquely defined. Let V− denote a generalized inverse of V (that
is a matrix with the property that VV− V = V). Then the random variable
yT V− y has the non-central chi-squared distribution with k degrees of free-
dom and non-centrality parameter λ = µ T V− µ .
For further details about properties of the chi-squared distribution see
Forbes et al. (2010).
9. Let y1 , . . . , yn be n independent random vectors each of length p and
yn ∼ MVN(0, V). Then S = ∑ni=i yi yTi is a p × p random matrix which
has the Wishart distribution W(V, n). This distribution can be used to
make inferences about the covariance matrix V because S is proportional
to V. In the case p = 1 the Yi ’s are independent random variables with
Yi ∼ N(0, σ 2 ), so Zi = Yi /σ ∼ N(0, 1). Hence, S = ∑ni=1 Yi2 = σ 2 ∑ni=1 Zi2
and therefore S/σ 2 ∼ χ 2 (n). Thus, the Wishart distribution can be re-
garded as a generalisation of the chi-squared distribution.

1.4.3 t-distribution
The t-distribution with n degrees of freedom is defined as the ratio of two
independent random variables. The numerator has the standard Normal distri-
bution and the denominator is the square root of a central chi-squared random
variable divided by its degrees of freedom; that is,
Z
T= (1.6)
(X 2 /n)1/2
where Z ∼ N(0, 1), X 2 ∼ χ 2 (n) and Z and X 2 are independent. This is denoted
by T ∼ t(n).

1.4.4 F-distribution
1. The central F-distribution with n and m degrees of freedom is defined
as the ratio of two independent central chi-squared random variables, each
QUADRATIC FORMS 11
divided by its degrees of freedom,

X2 X22
F= 1 , (1.7)
n m

where X12 ∼ χ 2 (n), X22 ∼ χ 2 (m) and X12 and X22 are independent. This is
denoted by F ∼ F(n, m).
2. The relationship between the t-distribution and the F-distribution can be
derived by squaring the terms in Equation (1.6) and using definition (1.7)
to obtain 
2 Z2 X 2
T = ∼ F(1, n) , (1.8)
1 n
that is, the square of a random variable with the t-distribution, t(n), has the
F-distribution, F(1, n).
3. The non-central F-distribution is defined as the ratio of two independent
random variables, each divided by its degrees of freedom, where the nu-
merator has a non-central chi-squared distribution and the denominator has
a central chi-squared distribution, that is,

X12 X22
F= ,
n m

where X12 ∼ χ 2 (n, λ ) with λ = µ T V−1 µ , X22 ∼ χ 2 (m), and X12 and X22 are
independent. The mean of a non-central F-distribution is larger than the
mean of central F-distribution with the same degrees of freedom.

1.4.5 Some relationships between distributions

We summarize the above relationships in Figure 1.1. In later chapters we add
to this diagram and a more extensive diagram involving most of the distri-
butions used in this book is given in the Appendix. Asymptotic relationships
are shown using dotted lines and transformations using solid lines. For more
details see Leemis (1986) from which this diagram was developed.

1.5 Quadratic forms

1. A quadratic form is a polynomial expression in which each term has de-
gree 2. Thus, y21 + y22 and 2y21 + y22 + 3y1 y2 are quadratic forms in y1 and y2 ,
but y21 + y22 + 2y1 or y21 + 3y22 + 2 are not.
12 INTRODUCTION

Standard X- Multivariate
Normal Normal n=1 Normal
N(0,1) X N( 2) MVN( 2)

n X21+...+Xn2

t Chi-square

t(n) 2(n)

X12/X22
n m
X2 m R=1
nX
F Wishart

F(n,m) W(R, n )

Figure 1.1 Some relationships between common distributions related to the Normal
distribution, adapted from Leemis (1986). Dotted line indicates an asymptotic rela-
tionship and solid lines a transformation.

2. Let A be a symmetric matrix

 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
 
 .. .. .. ,
 . . . 
an1 an2 · · · ann

where ai j = a ji ; then the expression yT Ay = ∑i ∑ j ai j yi y j is a quadratic

form in the yi ’s. The expression (y − µ )T V−1 (y − µ ) is a quadratic form
in the terms (yi − µi ) but not in the yi ’s.
3. The quadratic form yT Ay and the matrix A are said to be positive definite
if yT Ay > 0 whenever the elements of y are not all zero. A necessary and
sufficient condition for positive definiteness is that all the determinants
a11 a12 a13
a11 a12
|A1 | = a11 , |A2 | = , |A3 | = a21 a22 a23 , . . . , and
a21 a22
a31 a32 a33
|An | = det A are positive. If a matrix is positive definite, then it can be
inverted and also it has a square root matrix A∗ such that A∗ A = A. These
Exploring the Variety of Random
Documents with Different Content
in cattle kept on distillery and brewery dregs. Lead taken in small
quantities in soft water that has run through lead pipes or stood in
leaden cisterns produces in cows and other animals chronic
affections of the kidney. Ellenberger and Hofmeister have produced
the disease experimentally with lead and copper respectively.
Microbian invasions of the kidney that advance slowly like
glanders and tubercle are further causes of chronic nephritis. Other
secondary microbian infections of the kidney are complications of
infectious diseases in other parts, including abscess, pyæmia,
septicæmia, ulcerative endocarditis of the left heart, bronchitis,
pneumonia (Fröhner), and of others less directly in the line of the
circulation, as omphalitis, uterine phlebitis (Lustig), abscess of the
nasal sinuses, bones, and fistulæ (Trasbot).
In other cases the nephritis is evidently a result of the irritation
caused by toxins in process of elimination by the kidneys, as there is
no evidence of a nephritic infection.
In some instances minute emboli originating in the lungs or heart,
become the starting point of the nephritis, which slowly extends by
reason of infection or low condition and special susceptibility.
Disease of the aorta or renal artery may lead to this condition as
noticed by Cadeac and Lustig. Cadeac has also noticed its association
with aneurism of the mesenteric arteries so that the strongylus
(sclerostoma) armatus may be considered as a factor. Again in old
horses and dogs it has been associated with atheroma of the aorta
and renal vessels (Trasbot).
Overfeeding is not without its influence, especially when on animal
food, which charges the kidneys with excreting an excess of the
irritating urea and uric acid, and this is one reason why it is far more
frequent in house dogs than in other domestic animals. When the
meat is already decomposing and putrid there is the added evil of a
quantity of toxins and even of microbes to be eliminated from the
system by the much abused kidneys. Add to these that the dog’s
urine is even in the normal condition more dense and contains more
irritating ingredients than that of herbivora, and that owing to the
slight activity of his perspiratory apparatus he can obtain less relief
from the skin, and we find a substantial ground for the prevalence of
chronic nephritis in this animal.
 Disease of the valves of the right heart or dilatation with
insufficiency of the auriculo-ventricular valves is a potent cause of
nephritis, the reflux of blood into the veins and the increased venous
tension, speedily producing passive congestion and a slow type of
inflammation in the kidney. This factor is especially liable to operate
in dogs, which are particularly obnoxious to rheumatism and
valvular ulceration, and are very subject to nervous cardiac
disorders; in horses that have contracted heaves; and in beef breeds
of cattle which suffer from fatty degeneration of the heart with
dilatation.
The influence of calculi must not be overlooked, whether they are
lodged in the pelvis, the chalices, or the uriniferous tubules. Their
tendency is to induce local irritation and exudation, with fibroid
degeneration and thickening of the walls of the tubules or pelvis and
of the adjacent tissue.
When to one or more of the above conditions there are added
overfeeding or what is worse a low condition from starvation or
unwholesome food (permeated by bacteria or cryptogams or
containing vegetable acids), and when to crown all there are frequent
exposures to cold or wet, we have a vicious combination especially
conducive to kidney trouble.
Habitual retention of urine in mares in harness, in house dogs, or
in horses in railway cars, and violent exertion, or sprains of the back
are among the remaining accessory causes.
Symptoms. These are often slight or obscure, so that not only
owners and attendants but even veterinarians are liable to overlook
them. Loss of flesh, flabbiness of the muscles and a lack of spirit and
energy are among the first symptoms. The horse appears stiff,
especially in his loins and hind limbs, and fails to advance the hind
feet as far under the belly as formerly, and straddles more. When put
to work he is early fatigued and appears unfit for sustained exertion.
His movements are slow and if urged to a trot he may even groan
with every step and quickly settles back to his sluggish pace. If
turned sharply round on himself he does so with difficulty and often
groans. When he is mounted or when the loins are pinched he may
droop to excess. If you come on him lying down, and urge him to rise
he may rise on his fore limbs and sit on his haunches until urged
before he makes any attempt to raise himself on his hind. The dog
may spend most of his time in the kennel, and show little disposition
to run, play or hunt. On the contrary the owner may have to call him
several times before he will come out and then he moves listlessly,
wearily and even weakly.
In all animals the appetite is poor or capricious, and the patient
gradually loses condition, at first slowly and later, after a few weeks
or months, more rapidly. The advance of anæmia is also steadily
progressive.
Dropsical effusion is not uncommon. It is often prominent in the
horse as stocked limbs, but may be absent for a length of time. In
other animals it is more likely to appear later in the disease and
under the chest or abdomen or in one of the internal serous cavities.
Trasbot has found it absent for months in the nephritic dog.
The exploration of the kidney through the flaccid abdominal walls
in small animals, and through the rectum in small horses and cattle,
may reveal renal tenderness and even swelling. If there is a tendency
to frequent passage of urine in small quantities, or to straining
without micturition, the indication is of value.
There may be little or no fever, and, when left at rest, little
evidence of discomfort.
Any indication of urinary trouble, and especially with dropsy,
weakness, flabbiness and anæmia and a subnormal temperature,
should lead to examination of the urine, as a crucial test. A high
density is good ground for suspicion. But this is not constant. In
advanced cases (chronic interstitial nephritis, small white kidney,
atrophic nephritis) it may be 1015 to 1025, in exceptional advanced
cases with polyuria, it may be 1010, 1005, or even 1001. With such a
condition, however, there is great anæmia, pallor of the mucosæ, and
prostration. Tested with nitric acid and heat, the urine throws down
an abundant precipitate of albumen. Under the microscope it shows
a profusion of granular, degenerating epithelial cells, and casts of the
uriniferous tubes.
Progress. The course of the disease is usually slow, extending over
several months, but with a tendency to constant advance. The thirst
increases and the urine increases in amount, clearness and levity.
There may supervene extreme sluggishness, dropsies, anæmia, and
weakness, irritability of the heart, and palpitations on slight exertion.
So long as the heart’s action is strong, elimination may be
maintained and life prolonged for months (in cow, Dickinson), or
years (Friedberger and Fröhner). When the heart’s action becomes
weak, elimination is rendered imperfect and the animal shows
catarrh of the lungs or bowels (common in dogs), local inflammation
of the lungs, pleura or pericardium, or œdemas, or hæmorrhages.
The toxic effect on the nerve centres is shown by stupor or lethargy,
or vertigo. When an abscess forms it is associated with a temporary
rise of temperature (Trasbot). The patient may die in convulsions, in
a state of coma, or by gradually advancing debility and failure of the
heart.
Lesions. In cases of comparatively short standing the kidney is
usually of full size, or somewhat enlarged, with firmly adherent
capsule and rough or even nodular surface. The surface of the cortex
may be red or grayish or parti-colored, pink and gray. The cortical
portion is firm and it may even be attenuated somewhat, while the
medullary portion, naturally lighter, has often grayish streaks
converging toward the hilus. When the gray streaks are scraped with
the knife a serous fluid, mixed with fatty granules or globules, is
obtained. The glomeruli may be still about the normal size with some
increase of the epithelial tuft cells. The tubules contain casts (colloid,
hyaline, granular), and their epithelium normally columnar, are
flattened down to cubes and are swollen, granular or fatty.
In cases of older standing the connective tissue has usually
undergone a marked increase. The capsule is thick, dense and
adherent. The cortical substance is shrunken with a great increase of
the fibrous elements, and the same holds true of the medullary
portion. In consequence of this, even in the cortical substance the
white or gray color predominates. The parenchymatous tissue
(glomeruli, tubules) have greatly shrunken. In connection with the
contraction of the forming fibrous hyperplasia, there is a general
shrinkage of the kidney in size, it may be to one-half its original
volume. Trasbot reports a case of nephritis, of 8 months standing, in
the dog, with a kidney half the normal size. In the end the
parenchyma may have practically disappeared, and the kidney may
have shrunken to a small, firm, white, fibrous mass. Abscess of the
kidney is exceptionally met with (Laurent, Lafosse).
 Lesions of distant organs are not uncommon. Bronchitis,
pneumonia, pleurisy, insufficiency of the tricuspid or mitral valves,
dilated heart, hypertrophied or fatty heart, congested or fibroid liver,
arteritis, and dropsies are among such morbid conditions.
Prognosis. This is almost always unfavorable. Death may be
delayed for months or years, and partial transient recoveries may
take place but a restoration to normal structure and function is not to
be looked for.
Treatment. This cannot be expected to be much more than
palliative. The avoidance of overwork, and of the exposure to cold
and wet, and the securing of a free action of the skin by warm
buildings and clothing, are essential. The diet should be easily
digested and non-stimulating, for herbivora green food, carrots,
roots, apples, silage, with a moderate allowance of oats to counteract
weakness and anæmia; and for carnivora, milk, buttermilk, mush
made of oat, wheat or barley meal, with, if necessary, a slight
allowance of tender raw meat. Tonics fill a similar need. Iron and
bitters may be combined. Or hydrochloric acid or nitromuriatic acid
with bitters (nux, calumba, salicin, quassia) may be tried. These
acids are especially valuable when the case has originated in or is
maintained by calculi, indigestion or hepatic disorder. When the
heart is defective in tone, it may be stimulated by small doses of
digitalis, strophanthus, sparteine, caffein, or nitro-glycerine, or to a
certain extent by strychnia or nux. These, however, must be used
with judgment, if it is found that they aggravate the case by
increasing the arterial tension. In those cases in which there is an
excessive secretion of watery urine, the possible source of this in
musty aliment should be avoided, and the flow checked by nux
vomica, in moderate doses, and bromide or iodide of potassium in
full doses. When, on the other hand, the urine becomes scanty and
dense, the great danger of a toxic action must be met by agents that
favor excretion. Pure water at will is perhaps the least objectionable
of such agents, but potassium or sodium acetate or citrate, or even
sodium chloride, in weak solution, may be given. In some cases
benefit will come from a moderate use of the balsam of copiaba, or
the leaves of buchu, which may improve the tone of the secretory
elements. The most promptly effective of these agents is pilocarpin
(Friedberger and Fröhner), but it has the serious drawback of
inducing profuse and dangerous depletion and debility. Yet in careful
hands, and with good cardiac tone, it may often be used to
advantage.
Fomentations over the loins, warm baths and mustard
embrocations, may at times be beneficial. Attempts have been made
to check the hyperplasia by the use of arsenic, mercury or the
compounds of iodine, but their use in such cases is based on theory
rather than accomplished results.
 HYPERTROPHY OF THE KIDNEY.

Hypertrophy of both kidneys has not been recorded in domestic

animals. On the other hand the extraordinary development of one in
compensation for the loss or atrophy of the other is not uncommon.
In this the organ follows the general law of adaptation, seen in the
double symmetrical organs (testicle, etc.) and the more so that its
functional activity is indispensable to life. Among causes are:
blocking of an ureter by calculus, worms, neoplasm, nephritic
abscess, gangrene, etc. The enlargement of the remaining kidney is a
vicarious act and essentially a physiological one.
If compensation is perfect, it may be impossible to detect
symptoms apart from those of the primary disease.
Prognosis. Life is endangered in case of any subsequent kidney
disease.
 ATROPHY OF THE KIDNEY.

Result of hyperplasia of connective tissues and compression and absorption of

parenchyma. Unilateral or partial. Causes: chronic productive inflammation,
calculus in tubes, ureter, or pelvis, tumor, retention cyst, embolism. Lesions:
sclerosis of kidney, firmness, pallor, anæmia, lack of glomeruli and tubules, cysts,
congenital, urinous retention, colloid. Symptoms: reduced secretion, palpation of
kidney. Treatment: Prevention: arrest conditions, abundance of water, succulent
food, parasiticides, operation on cysts, counteract nephritis.

Unlike hypertrophy, this is constantly the result of a pathological

process. So long as a normal functional activity of the secreting
elements is carried on, such parts must maintain their size and
healthy characters. But with the compression of such secreting
elements (glomeruli and convoluted tubes) by a hyperplasia of
connective tissue, by pressure from without or from the damming
back of the urine in the pelvis and tubes, the secretory elements are
absorbed and removed, and the final result is a general atrophy. If
such atrophy appears in both kidneys at once it can only be very
partial in extent, as extreme atrophy of both, with loss of their
secretory function, would entail poisoning and death from the
retained urinary products. The comparative frequency of the disease
may be inferred from the reports of the numbers of specimens found
by Barrier and Moussu in old horses in the dissecting rooms. The
latter observed a dozen cases in a single winter, other examples are
recorded by Cadeac (horse), Soula (swine) and Trasbot (in various
animals).
Causes. The most common source of the condition is the
occurrence of chronic productive inflammation. The new product in
such cases, if not pus, or a growth that rapidly passes into fatty or
granular degeneration, or into gangrene, tends to form tissue of a
low organization, especially fibrous. The resulting increase of the
fibrous trabeculæ, in undergoing subsequent contraction necessarily
compresses the secretory tissue and the final result is a visible and, it
may be, extreme wasting. Hence any slowly advancing productive
inflammation is liable to result in absorption and removal of the
kidney parenchyma, and distinct atrophy of the gland.
Again the obstruction of the ureter by a calculus in the pelvis
which falls into the infundibuliform entrance, or a stone arrested at
any part of the duct (or even of the urethra) or by worms, hydatids,
cysts or tumors, throws back on the kidney the secreted urine, which
distending the pelvis and uriniferous tubes leads to direct
compression and absorption of the secretory parenchyma. Direct
compression of the kidney by an adjacent tumor will act in a similar
manner. Retention cysts by their gradual increase and augmenting
pressure cause absorption of the gland tissue.
The blocking of individual uriniferous tubules by minute calculi,
which is so often seen in cattle, kept on dry feeding in winter, is a
cause of partial nephritis, and absorption, as noted by Röll.
A somewhat rare cause of atrophy is the diminution of the blood
supply by arteritis and embolism of the renal artery, or by pressure
of tumors on that vessel. Arteritis and blocking suggests at once the
possible agency of the strongylus (sclerostoma) armatus in the horse.
Trasbot records a striking instance of compression of the renal artery
and kidney by an enormous sublumbar melanoma. This occurred in
an aged horse and led to atrophy.
Lesions. In cases due to productive inflammation with sclerosis of
the kidney, the firmness, pallor and bloodlessness of the organ is a
marked feature. When incised it is found to be composed mainly of
fibrous tissue, while the glomeruli and tubuli have to a large extent
disappeared.
If there has been simple lack of circulation the kidney becomes
flaccid, pale and small in size. The secretory elements (glomeruli and
uriniferous tubes) are first absorbed, leaving the fibrous network,
which tends to shrink and form a hard resistent mass. In extreme
cases there may be absolutely no glandular tissue left, and the dense
shrunken mass represents only the hyperplasia of the original
fibrous network. In the different successive stages of this process the
glomeruli and tubules become flattened, the epithelial cells become
granular, or contain colloid casts and refrangent elements like oil
globules and finally they are represented by a small mass of fibrous
material.
Of all the atrophies caused by the pressure of tumors perhaps that
caused by cysts is the most characteristic. There may be a single cyst
or they may be multiple; they may range in size from a pea to the size
of the two fists the total size exceeding that of the normal kidney. In
all such cases the cysts project visibly from the surface of the organ.
They vary according to their origin and nature. Congenital cysts are
said to have resulted from distension by retained urine of the capsule
of the glomerulus. The arterial tuft is atrophied and flattened against
the wall. Serous cysts with clear contents are found in the old.
Urinous cysts again form by distension of the tubules that are
obstructed by cysts or minute calculi. Colloid cysts are found in
certain forms of nephritis formed by the dilatation of the capsule of
the glomerulus or of the uriniferous tubules. The liquid often
contains leucin, tyrosin and cholesterine. In all such cases the walls
of the cyst become thick, and the glandular parenchyma is
compressed leading to progressive degeneration and atrophy.
Symptoms of atrophy of the kidney are necessarily those of
suppression of urine, with, in certain cases, the passage of casts of
the uriniferous tubes and of crystals of salts. There are, however, no
absolutely pathognomonic symptoms. When the kidney can be
reached through the flaccid walls of a comparatively empty
abdomen, or through the rectum, its hard, shrunken condition may
assist in diagnosis.
Treatment is not successful in advanced cases. Prevention is to be
sought by obviating or treating the conditions on which the atrophy
depends. Nephritis must be treated on general principles. Calculi
must be avoided by a liberal supply of water, by soiling, or by
pasturage. Strongylus parasitism should be dealt with by destroying
the parent worms in the bowels, and by securing pure drinking water
free from their eggs and embryos. Cysts, and tumors are only
amenable to surgical measures and not often open even to these.
 FATTY DEGENERATION OF THE KIDNEY:
STEATOSIS OF THE KIDNEY.

Causes: age, overfeeding, idleness, atony, retention of urine. Lesions: kidney

enlarged, pale yellow, capsule loose, cut surface glistening unctuous, oil globules in
scrapings, granules soluble in ether. Symptoms: in idle, overfed, obese, improved
meat producing breeds, closely confined, starchy or saccharine food, fatty granules
in urine, finally dropsies, anæmia, debility, sluggishness. Prognosis unfavorable in
advanced stage. Treatment: butcher, restricted regimen, open air exercise,
nitrogenous diet, crossing, diuretic food or drugs, oil of turpentine, balsam
copiaba. Palliation only.

Fatty degeneration of the kidneys is by no means unknown in the

domestic animals. It has been observed in dogs and cats (Rogers,
Goubaux, Vulpain, Trasbot). In dogs it has been erroneously set
down as a characteristic lesion of rabies. Like fatty degeneration of
other organs, it is also met with in old and overfed individuals of
meat producing breeds of animals, in which the tendency to early
maturity and rapid and excessive fattening has been fostered from
generation to generation. In man small, granular, fatty kidney is a
common result of chronic parenchymatous nephritis, and often
coincides with fatty liver. Chronic poisoning by arsenic or
phosphorus is another cause, as it is of fatty degeneration in other
organs.
Vulpain has attributed it to a lack of active exertion and of general
tone, associated with excessive amylaceous feeding, sluggish, shallow
breathing and tardy elimination. Goubaux and Trasbot attach great
importance to the compulsory retention of urine in house dogs, cats
and horses. The damming back of the urine in the convoluted tubes
and glomeruli, temporarily arrests secretion, and the inactive and
compressed cells tend at once to granular and fatty degeneration.
 Lesions. The gland is sensibly increased in size, and pale, yellowish
or straw yellow. The capsule is easily detached from the cortical
substance, contrary to what is the case in chronic productive
inflammation. The cortical substance is increased in thickness, and
pale, the pallor being largely in ratio with the duration or extent of
the fatty degeneration. The cut surface may be glistening and
unctuous to the touch. It is softer than usual, rather friable, and if
scraped, furnishes a serous or grayish pulp in which oil globules are
prominent features, together with granular epithelium and free
granules that dissolve readily in ether. Tubules are varicose and
unequal at different parts. The medullary portion has undergone
little change. It may be paler at certain points, with some shrinking
of its substance and increase of firmness.
Symptoms. As a rule the disease occurs in pampered, overfed and
obese animals, and in those of the improved breeds which have great
power of digestion, assimilation and fattening. It is especially to be
looked for after close confinement on full, stimulating, amylaceous
diet. Symptoms are not usually recognized during life. There is,
however, a lessening of the urinary secretion, and, as the disease
advances, albuminuria. When examined microscopically this is
found to contain characteristic elements, such as granular epithelial
cells, the granules soluble in ether, oil globules, and at times crystals
of cholesterine (Beale). A diagnosis based on the mere presence of oil
globules may, however, be fallacious, as these may be present in
animals that have just been heavily fed on oleaginous food, and again
the oil used to smear the catheter may float in the urine and prove
misleading. Under such circumstances vaseline or glycerine may be
substituted on the catheter. Scriba induced fatty urine by injecting
fat or oil emulsion into the blood, and Chabrie by ligating the large
intestine. Trasbot says that cylindroid casts may be present. As in
other grave kidney affections, dropsies supervene as the disease
advances. These may show in the limbs, in the abdomen, or in other
serous cavities. A steadily advancing anæmia with pallor of the
mucosæ, listlessness, weakness, debility and sluggishness are to be
noted.
Prognosis. Since the disease is rarely diagnosed until it has
reached an advanced stage, it usually progresses steadily to a fatal
issue. If, however, it can be detected at an earlier stage, it may be
palliated, or held in abeyance, for a length of time varying with the
extent of the lesions. As it is very largely a disease of meat producing
animals and as the subject is at first in a condition of marked obesity,
it can usually be turned over to the butcher without material loss.
Treatment. If the disease has resulted from the inbred propensity
to fattening, the family that shows the disposition must be subjected
to a somewhat different regimen, open air exercise must take the
place of confinement in warm stables, a rather bare pasturage is
valuable for herbivora, and a restricted diet in which the oleaginous,
saccharine, and amylaceous constituents do not predominate, is
strongly indicated. Crossing with a strange male having many of the
desirable qualities of the herd, but which is more vigorous may be
resorted to. When the secretion of urine becomes scanty an
abundance of pure water, or a diet of succulent grass or roots or
ensilage or even small doses of alkaline diuretics may be resorted to.
Any source of arsenic or phosphorus poisoning should be cut off, and
as an antidote to phosphorus, oil of turpentine may be given in small
doses. This agent may, indeed, replace the alkalies as a diuretic,
bringing in an element of tone for the mucosa which is not to be
despised. Or balsam of copaiba or buchu leaves may be substituted.
When the small white kidney (granular, fatty) results from chronic
nephritis, the prevention and treatment would be as for that disease.
Little hope is to be entertained of entire restoration to health.
 AMYLOID KIDNEY. LARDACEOUS OR WAXY
KIDNEY.

This condition of the kidney has been found in the ox (Gerlach)

and dog (Rabe, etc.). There are usually similar degenerative lesions
in the liver, pancreas, intestines and other organs. It is usually a
concomitant of some chronic wasting disease (chronic nephritis,
tuberculosis, etc.).
Morbid Anatomy. The kidney is usually enlarged, pale and on
section waxy or glistening. Soaked in dilute compound tincture of
iodine it shows spots of a walnut or mahogany brown color. The
glomeruli are well marked and show the earlier changes, later the
tubes do so excepting the epithelium. The latter is swollen, granular,
fatty.
Symptoms. There may have been those of chronic nephritis. Rabe
has noticed in dogs dropsy of the limbs, ascites, emaciation,
anorexia, followed by uræmia, coma, weakness, vomiting, and if the
kidney alone was affected great lowering of temperature (35.9°C).
With hepatic complication there was greater weakness, giddiness,
and higher temperature (39.6°C). Urine is usually increased (in man
albuminous) and the casts have shown the anyloid reaction. They
tend to be fatty or finely granular. Casts may, however, show anyloid
reaction when the kidney, post mortem, does not (Jaksch).
Diagnosis from Bright’s disease is often impossible.
Treatment is essentially the same as in chronic nephritis, and is
not hopeful.
Trasbot recommends KI 3 to 7 grs., or tinct of iodine 3 drops for
shepherd dog. Ol. terebinth and alkaline diuretics are also
commended.
 RENAL CALCULUS.

This is much more common than is supposed. Small calculi formed

in the tubuli uriniferi of cattle on dry winter feeding often pass
without recognition, and habitually disappear on rich spring and
summer grass.
If retained in the pelvis until increasing size forbids their passage
through the ureter they form pelvic calculi.
If retained in the bladder so that they cannot enter the urethra
they form cystic calculi.
Pelvic calculi or concretions are often (in cattle and swine) mere
scales lying in chalices. They may fill the whole pelvis and send
branching processes into chalices.
Causes. They are attributed to phosphaturia, lithæmia or uric acid
diathesis, oxaluria, etc. In cattle they are associated with dry feeding
and are common on all magnesian limestone soils. There are usually
catarrh of the kidney and the presence of bacterial ferments and
colloids (pus, albumen, etc.). (Sharing and Ord.) Calculi or gravel is
preceded by renal catarrh, but this is aggravated by the crystalline
deposit. Bacteria act also in producing NH3O, which instantly
precipitates ammonio-magnesian phosphate. Retention of urine
greatly favors the precipitation.
Symptoms. A white or brownish yellow deposit in the last urine
discharged collects on the floor. Cloudy urine. Passage of crystals—
round—or angular. Colic. Lameness in one or both hind limbs.
Arched back. Sensitive loins. Pain paroxysmal. Attempts to urinate.
Little passed but often with drops of blood. Sudden relief when the
calculus enters the bladder.
Retained in the kidney it may cause no suffering in meat
producing animals, but in horses it usually causes stiffness or
lameness especially under violent effort. Also hematuria; blood
globules are found in the deposit when placed under the microscope.
There may be sepsis and specially cloudy offensive urine.
Diagnosis: May be confounded with renal tuberculosis, or sarcoma
or oxaluria. Examine for bacillus, small cells, or oxalate of lime or
oxalic acid.
Prophylaxis. In the early stages give succulent, watery food,
ensilage, roots, potatoes, spring grass, and water ad libitum.
Treatment. Salt may tempt the patient to drink. Nitro-muriatic
acid is a solvent and antiseptic. Or alkalies with salicylate of soda.
Also tonics. Quiet pain by morphia and other anodynes. Use
piperazine.
These failing, an operation on the kidney may be considered.
 HYDRO-NEPHROSIS.
A common result of calculus or other obstruction, causing
increasing pressure of urine in the pelvis and absorption of the
parenchyma, and finally leaving a mere urinous sac.
 RENAL TUMORS.

1. Non-malignant: Fibroma.
Lipoma.
Angioma.
Adenoma.
Papilloma.
2. Malignant: Sarcoma.
Carcinoma.
 RENAL PARASITES.
Echinococcus: Herbivora, Omnivora.
Bilharzia Crassa: Egyptian cattle.
Strongylus Gigas: Horse, ox, dog, man.
(Cysticercus Tenuicollis: Ruminants: Pig).
Tænia serrata: Dog. Pelvis.
Sclerostoma equinum: (renal arteries, kidney pelvis), soliped.
Stephanurus dentatus: Pig, (pus cavities).
Trichosoma plicata: (Urinary bladder), dog.
T. felis: (Cat), bladder.
Indetermined embryos: Kidneys, dog; small tumors.
Cytodites nudus: Kidneys; hens.
Œstrus, (Gast. Hemorrhoidalis): Bladder walls: horse.
Mucorimyces: Kidneys; dog.
Coccidia: Kidney, Horse, dog, goose.
 INJURIES OF THE URETERS.

Lesions by bullets, arrows, stabs, bruises and lacerations in parturition, treads,

wheels, tumors, ulcers, calculi, tubercles, parasites. Course. Pathology: transverse
division may cause hydronephrosis, or septic peritonitis. Symptoms: uncertain,
traumatism, bloody urine, arched, stiff, tender loins, straining, recumbency,
groaning in turning or rising, rectal palpation of distended ureter, of ascitic fluid,
pitting on pressure of loin, flank or groin, liquid drawn through a cannula is
urinous, urine still discharged by normal channel. Crystals in urine, worm ova.
Treatment: compresses, fomentations, sinapisms, anodynes, balsams,
antispasmodics, extraction of calculus, lateral implantation of urethra.

From their deep and protected position it might be plausibly

concluded that the ureters were secure against every kind of
traumatism. This however, is not the case, since in both man and
animals they have been known to have been injured by bullet
wounds, arrow wounds, and stab wounds of various kinds. In
dystokia with laceration of the womb, vagina or bladder the ureter is
liable to be injured. By blows and kicks it may even be ruptured or
torn across, and also by sudden and severe mechanical compression
of the abdomen as when run over by a wagon or trodden on by a
horse, ox, or other large animal. Tumors of various kinds may grow
in, or press upon the ureter, ulcers with thick indurated margins or
base may obstruct the passage, or calculi, or worms may block and
give rise to overdistension and even rupture. Kopp describes
obstruction by multiple calculi with saccular dilation in front, close
to the kidney in a cow. Cadiot records cases of thickening of the
mucosa by numerous cysts as small as hempseed. Intra-abdominal
tumors of the spermatic cord have been known to block the passage.
Again tubercles have formed on the urethra, and polypi on the
trigonum vesicæ have blocked the ureter and produced all the evil
consequences of calculus, parasites, etc.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com