Statistical Analysis

of Spatial and Spatio-

Temporal Point
Patterns
Third Edition
 MONOGRAPHS ON STATISTICS AND APPLIED PROBABILITY

General Editors

F. Bunea, V. Isham, N. Keiding, T. Louis, R. L. Smith, and H. Tong

1. Stochastic Population Models in Ecology and Epidemiology M.S. Barlett (1960)

2. Queues D.R. Cox and W.L. Smith (1961)
3. Monte Carlo Methods J.M. Hammersley and D.C. Handscomb (1964)
4. The Statistical Analysis of Series of Events D.R. Cox and P.A.W. Lewis (1966)
5. Population Genetics W.J. Ewens (1969)
6. Probability, Statistics and Time M.S. Barlett (1975)
7. Statistical Inference S.D. Silvey (1975)
8. The Analysis of Contingency Tables B.S. Everitt (1977)
9. Multivariate Analysis in Behavioural Research A.E. Maxwell (1977)
10. Stochastic Abundance Models S. Engen (1978)
11. Some Basic Theory for Statistical Inference E.J.G. Pitman (1979)
12. Point Processes D.R. Cox and V. Isham (1980)
13. Identification of Outliers D.M. Hawkins (1980)
14. Optimal Design S.D. Silvey (1980)
15. Finite Mixture Distributions B.S. Everitt and D.J. Hand (1981)
16. Classification A.D. Gordon (1981)
17. Distribution-Free Statistical Methods, 2nd edition J.S. Maritz (1995)
18. Residuals and Influence in Regression R.D. Cook and S. Weisberg (1982)
19. Applications of Queueing Theory, 2nd edition G.F. Newell (1982)
20. Risk Theory, 3rd edition R.E. Beard, T. Pentikäinen and E. Pesonen (1984)
21. Analysis of Survival Data D.R. Cox and D. Oakes (1984)
22. An Introduction to Latent Variable Models B.S. Everitt (1984)
23. Bandit Problems D.A. Berry and B. Fristedt (1985)
24. Stochastic Modelling and Control M.H.A. Davis and R. Vinter (1985)
25. The Statistical Analysis of Composition Data J. Aitchison (1986)
26. Density Estimation for Statistics and Data Analysis B.W. Silverman (1986)
27. Regression Analysis with Applications G.B. Wetherill (1986)
28. Sequential Methods in Statistics, 3rd edition G.B. Wetherill and K.D. Glazebrook (1986)
29. Tensor Methods in Statistics P. McCullagh (1987)
30. Transformation and Weighting in Regression R.J. Carroll and D. Ruppert (1988)
31. Asymptotic Techniques for Use in Statistics O.E. Bandorff-Nielsen and D.R. Cox (1989)
32. Analysis of Binary Data, 2nd edition D.R. Cox and E.J. Snell (1989)
33. Analysis of Infectious Disease Data N.G. Becker (1989)
34. Design and Analysis of Cross-Over Trials B. Jones and M.G. Kenward (1989)
35. Empirical Bayes Methods, 2nd edition J.S. Maritz and T. Lwin (1989)
36. Symmetric Multivariate and Related Distributions K.T. Fang, S. Kotz and K.W. Ng (1990)
37. Generalized Linear Models, 2nd edition P. McCullagh and J.A. Nelder (1989)
38. Cyclic and Computer Generated Designs, 2nd edition J.A. John and E.R. Williams (1995)
39. Analog Estimation Methods in Econometrics C.F. Manski (1988)
40. Subset Selection in Regression A.J. Miller (1990)
41. Analysis of Repeated Measures M.J. Crowder and D.J. Hand (1990)
42. Statistical Reasoning with Imprecise Probabilities P. Walley (1991)
43. Generalized Additive Models T.J. Hastie and R.J. Tibshirani (1990)
44. Inspection Errors for Attributes in Quality Control N.L. Johnson, S. Kotz and X. Wu (1991)
45. The Analysis of Contingency Tables, 2nd edition B.S. Everitt (1992)
46. The Analysis of Quantal Response Data B.J.T. Morgan (1992)
47. Longitudinal Data with Serial Correlation—A State-Space Approach R.H. Jones (1993)
48. Differential Geometry and Statistics M.K. Murray and J.W. Rice (1993)
49. Markov Models and Optimization M.H.A. Davis (1993)
50. Networks and Chaos—Statistical and Probabilistic Aspects
O.E. Barndorff-Nielsen, J.L. Jensen and W.S. Kendall (1993)
51. Number-Theoretic Methods in Statistics K.-T. Fang and Y. Wang (1994)
52. Inference and Asymptotics O.E. Barndorff-Nielsen and D.R. Cox (1994)
53. Practical Risk Theory for Actuaries C.D. Daykin, T. Pentikäinen and M. Pesonen (1994)
54. Biplots J.C. Gower and D.J. Hand (1996)
55. Predictive Inference—An Introduction S. Geisser (1993)
56. Model-Free Curve Estimation M.E. Tarter and M.D. Lock (1993)
57. An Introduction to the Bootstrap B. Efron and R.J. Tibshirani (1993)
58. Nonparametric Regression and Generalized Linear Models P.J. Green and B.W. Silverman (1994)
59. Multidimensional Scaling T.F. Cox and M.A.A. Cox (1994)
60. Kernel Smoothing M.P. Wand and M.C. Jones (1995)
61. Statistics for Long Memory Processes J. Beran (1995)
62. Nonlinear Models for Repeated Measurement Data M. Davidian and D.M. Giltinan (1995)
63. Measurement Error in Nonlinear Models R.J. Carroll, D. Rupert and L.A. Stefanski (1995)
64. Analyzing and Modeling Rank Data J.J. Marden (1995)
65. Time Series Models—In Econometrics, Finance and Other Fields
D.R. Cox, D.V. Hinkley and O.E. Barndorff-Nielsen (1996)
66. Local Polynomial Modeling and its Applications J. Fan and I. Gijbels (1996)
67. Multivariate Dependencies—Models, Analysis and Interpretation D.R. Cox and N. Wermuth (1996)
68. Statistical Inference—Based on the Likelihood A. Azzalini (1996)
69. Bayes and Empirical Bayes Methods for Data Analysis B.P. Carlin and T.A Louis (1996)
70. Hidden Markov and Other Models for Discrete-Valued Time Series I.L. MacDonald and W. Zucchini (1997)
71. Statistical Evidence—A Likelihood Paradigm R. Royall (1997)
72. Analysis of Incomplete Multivariate Data J.L. Schafer (1997)
73. Multivariate Models and Dependence Concepts H. Joe (1997)
74. Theory of Sample Surveys M.E. Thompson (1997)
75. Retrial Queues G. Falin and J.G.C. Templeton (1997)
76. Theory of Dispersion Models B. Jørgensen (1997)
77. Mixed Poisson Processes J. Grandell (1997)
78. Variance Components Estimation—Mixed Models, Methodologies and Applications P.S.R.S. Rao (1997)
79. Bayesian Methods for Finite Population Sampling G. Meeden and M. Ghosh (1997)
80. Stochastic Geometry—Likelihood and computation
O.E. Barndorff-Nielsen, W.S. Kendall and M.N.M. van Lieshout (1998)
81. Computer-Assisted Analysis of Mixtures and Applications—Meta-Analysis, Disease Mapping and Others
D. Böhning (1999)
82. Classification, 2nd edition A.D. Gordon (1999)
83. Semimartingales and their Statistical Inference B.L.S. Prakasa Rao (1999)
84. Statistical Aspects of BSE and vCJD—Models for Epidemics C.A. Donnelly and N.M. Ferguson (1999)
85. Set-Indexed Martingales G. Ivanoff and E. Merzbach (2000)
86. The Theory of the Design of Experiments D.R. Cox and N. Reid (2000)
87. Complex Stochastic Systems O.E. Barndorff-Nielsen, D.R. Cox and C. Klüppelberg (2001)
88. Multidimensional Scaling, 2nd edition T.F. Cox and M.A.A. Cox (2001)
89. Algebraic Statistics—Computational Commutative Algebra in Statistics
G. Pistone, E. Riccomagno and H.P. Wynn (2001)
90. Analysis of Time Series Structure—SSA and Related Techniques
N. Golyandina, V. Nekrutkin and A.A. Zhigljavsky (2001)
91. Subjective Probability Models for Lifetimes Fabio Spizzichino (2001)
92. Empirical Likelihood Art B. Owen (2001)
93. Statistics in the 21st Century Adrian E. Raftery, Martin A. Tanner, and Martin T. Wells (2001)
94. Accelerated Life Models: Modeling and Statistical Analysis
Vilijandas Bagdonavicius and Mikhail Nikulin (2001)
95. Subset Selection in Regression, Second Edition Alan Miller (2002)
96. Topics in Modelling of Clustered Data Marc Aerts, Helena Geys, Geert Molenberghs, and Louise M. Ryan (2002)
97. Components of Variance D.R. Cox and P.J. Solomon (2002)
98. Design and Analysis of Cross-Over Trials, 2nd Edition Byron Jones and Michael G. Kenward (2003)
99. Extreme Values in Finance, Telecommunications, and the Environment
Bärbel Finkenstädt and Holger Rootzén (2003)
100. Statistical Inference and Simulation for Spatial Point Processes
Jesper Møller and Rasmus Plenge Waagepetersen (2004)
101. Hierarchical Modeling and Analysis for Spatial Data
Sudipto Banerjee, Bradley P. Carlin, and Alan E. Gelfand (2004)
102. Diagnostic Checks in Time Series Wai Keung Li (2004)
103. Stereology for Statisticians Adrian Baddeley and Eva B. Vedel Jensen (2004)
104. Gaussian Markov Random Fields: Theory and Applications Håvard Rue and Leonhard Held (2005)
105. Measurement Error in Nonlinear Models: A Modern Perspective, Second Edition
Raymond J. Carroll, David Ruppert, Leonard A. Stefanski, and Ciprian M. Crainiceanu (2006)
106. Generalized Linear Models with Random Effects: Unified Analysis via H-likelihood
Youngjo Lee, John A. Nelder, and Yudi Pawitan (2006)
107. Statistical Methods for Spatio-Temporal Systems
Bärbel Finkenstädt, Leonhard Held, and Valerie Isham (2007)
108. Nonlinear Time Series: Semiparametric and Nonparametric Methods Jiti Gao (2007)
109. Missing Data in Longitudinal Studies: Strategies for Bayesian Modeling and Sensitivity Analysis
Michael J. Daniels and Joseph W. Hogan (2008)
110. Hidden Markov Models for Time Series: An Introduction Using R
Walter Zucchini and Iain L. MacDonald (2009)
111. ROC Curves for Continuous Data Wojtek J. Krzanowski and David J. Hand (2009)
112. Antedependence Models for Longitudinal Data Dale L. Zimmerman and Vicente A. Núñez-Antón (2009)
113. Mixed Effects Models for Complex Data Lang Wu (2010)
114. Intoduction to Time Series Modeling Genshiro Kitagawa (2010)
115. Expansions and Asymptotics for Statistics Christopher G. Small (2010)
116. Statistical Inference: An Integrated Bayesian/Likelihood Approach Murray Aitkin (2010)
117. Circular and Linear Regression: Fitting Circles and Lines by Least Squares Nikolai Chernov (2010)
118. Simultaneous Inference in Regression Wei Liu (2010)
119. Robust Nonparametric Statistical Methods, Second Edition
Thomas P. Hettmansperger and Joseph W. McKean (2011)
120. Statistical Inference: The Minimum Distance Approach
Ayanendranath Basu, Hiroyuki Shioya, and Chanseok Park (2011)
121. Smoothing Splines: Methods and Applications Yuedong Wang (2011)
122. Extreme Value Methods with Applications to Finance Serguei Y. Novak (2012)
123. Dynamic Prediction in Clinical Survival Analysis Hans C. van Houwelingen and Hein Putter (2012)
124. Statistical Methods for Stochastic Differential Equations
Mathieu Kessler, Alexander Lindner, and Michael Sørensen (2012)
125. Maximum Likelihood Estimation for Sample Surveys
R. L. Chambers, D. G. Steel, Suojin Wang, and A. H. Welsh (2012)
126. Mean Field Simulation for Monte Carlo Integration Pierre Del Moral (2013)
127. Analysis of Variance for Functional Data Jin-Ting Zhang (2013)
128. Statistical Analysis of Spatial and Spatio-Temporal Point Patterns, Third Edition Peter J. Diggle (2013)
Monographs on Statistics and Applied Probability 128

Statistical Analysis
of Spatial and Spatio-
Temporal Point
Patterns
Third Edition

Peter J. Diggle
Lancaster University
England, UK
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2014 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

Version Date: 20130625

International Standard Book Number-13: 978-1-4665-6024-6 (eBook - PDF)

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 stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copy-
right.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 pro-
vides licenses and registration for a variety of users. For organizations that have been granted a pho-
tocopy 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.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To the memory of Julian Besag FRS, 1945-2010
This page intentionally left blank
Contents

List of Figures xv

List of Tables xxvii

Preface xxix

1 Introduction 1

1.1 Spatial point patterns . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Edge-effects . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Complete spatial randomness . . . . . . . . . . . . . . . . . . 9
1.5 Objectives of statistical analysis . . . . . . . . . . . . . . . . 11
1.6 The Dirichlet tessellation . . . . . . . . . . . . . . . . . . . . 12
1.7 Monte Carlo tests . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2 Preliminary testing 17

2.1 Tests of complete spatial randomness . . . . . . . . . . . . . 17

2.2 Inter-event distances . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 Analysis of Japanese black pine saplings . . . . . . . . 21
2.2.2 Analysis of redwood seedlings . . . . . . . . . . . . . . 21
2.2.3 Analysis of biological cells . . . . . . . . . . . . . . . . 22
2.2.4 Small distances . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Nearest neighbour distances . . . . . . . . . . . . . . . . . . 24
2.3.1 Analysis of Japanese black pine saplings . . . . . . . . 25
2.3.2 Analysis of redwood seedlings . . . . . . . . . . . . . . 25
2.3.3 Analysis of biological cells . . . . . . . . . . . . . . . . 26
2.4 Point to nearest event distances . . . . . . . . . . . . . . . . 26
2.4.1 Analysis of Japanese black pine seedlings . . . . . . . 28
2.4.2 Analysis of redwood seedlings . . . . . . . . . . . . . . 28
2.4.3 Analysis of biological cells . . . . . . . . . . . . . . . . 28
2.5 Quadrat counts . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.1 Analysis of Japanese black pine seedlings . . . . . . . 30
2.5.2 Analysis of redwood seedlings . . . . . . . . . . . . . . 31
2.5.3 Analysis of biological cells . . . . . . . . . . . . . . . . 32

ix
x Contents

2.6 Scales of pattern . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.6.1 Analysis of Lansing Woods data . . . . . . . . . . . . 33
2.6.2 Scales of dependence . . . . . . . . . . . . . . . . . . . 34
2.7 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . 35

3 Methods for sparsely sampled patterns 39

3.1 General remarks . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2 Quadrat counts . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1 Tests of CSR . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Estimators of intensity . . . . . . . . . . . . . . . . . . 42
3.2.3 Analysis of Lansing Woods data . . . . . . . . . . . . 42
3.3 Distance measurements . . . . . . . . . . . . . . . . . . . . . 43
3.3.1 Distribution theory under CSR . . . . . . . . . . . . . 44
3.3.2 Tests of CSR . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.3 Estimators of intensity . . . . . . . . . . . . . . . . . . 49
3.3.4 Analysis of Lansing Woods data . . . . . . . . . . . . 50
3.3.5 Catana’s wandering quarter . . . . . . . . . . . . . . . 51
3.4 Tests of independence . . . . . . . . . . . . . . . . . . . . . . 52
3.5 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . 53

4 Spatial point processes 55

4.1 Processes and summary descriptions . . . . . . . . . . . . . . 55

4.2 Second-order properties . . . . . . . . . . . . . . . . . . . . . 57
4.2.1 Univariate processes . . . . . . . . . . . . . . . . . . . 57
4.2.2 Extension to multivariate processes . . . . . . . . . . . 60
4.3 Higher order moments and nearest neighbour distributions . 60
4.4 The homogeneous Poisson process . . . . . . . . . . . . . . . 61
4.5 Independence and random labelling . . . . . . . . . . . . . . 63
4.6 Estimation of second-order properties . . . . . . . . . . . . . 64
4.6.1 Stationary processes . . . . . . . . . . . . . . . . . . . 64
4.6.2 Estimating the pair correlation function . . . . . . . . 71
4.6.3 Intensity-reweighted stationary processes . . . . . . . 72
4.6.4 Multivariate processes . . . . . . . . . . . . . . . . . . 73
4.6.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.7 Displaced amacrine cells in the retina of a rabbit . . . . . . . 76
4.8 Estimation of nearest neighbour distributions . . . . . . . . . 78
4.8.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.9 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . 79

5 Nonparametric methods 83

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Estimating weighted integrals of the second-order intensity . 83
Contents xi

5.3 Nonparametric estimation of a spatially varying intensity . . 84

5.3.1 Estimating spatially varying intensities for the Lansing
Woods data . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Analysing replicated spatial point patterns . . . . . . . . . . 90
5.4.1 Estimating the K-function from replicated data . . . . 92
5.4.2 Between-group comparisons in designed experiments . 94
5.5 Parametric or nonparametric methods? . . . . . . . . . . . . 97

6 Models 99

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.2 Contagious distributions . . . . . . . . . . . . . . . . . . . . 100
6.3 Poisson cluster processes . . . . . . . . . . . . . . . . . . . . 101
6.4 Inhomogeneous Poisson processes . . . . . . . . . . . . . . . 104
6.5 Cox processes . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.6 Trans-Gaussian Cox processes . . . . . . . . . . . . . . . . . 109
6.7 Simple inhibition processes . . . . . . . . . . . . . . . . . . . 110
6.8 Markov point processes . . . . . . . . . . . . . . . . . . . . . 112
6.8.1 Pairwise interaction point processes . . . . . . . . . . 114
6.8.2 More general forms of interaction . . . . . . . . . . . . 118
6.9 Other constructions . . . . . . . . . . . . . . . . . . . . . . . 118
6.9.1 Lattice-based processes . . . . . . . . . . . . . . . . . 118
6.9.2 Thinned processes . . . . . . . . . . . . . . . . . . . . 119
6.9.3 Superpositions . . . . . . . . . . . . . . . . . . . . . . 120
6.9.4 Interactions in an inhomogeneous environment . . . . 121
6.10 Multivariate models . . . . . . . . . . . . . . . . . . . . . . . 123
6.10.1 Marked point processes . . . . . . . . . . . . . . . . . 123
6.10.2 Multivariate point processes . . . . . . . . . . . . . . . 123
6.10.3 How should multivariate models be formulated? . . . . 124
6.10.4 Cox processes . . . . . . . . . . . . . . . . . . . . . . . 125
6.10.5 Markov point processes . . . . . . . . . . . . . . . . . 128

7 Model-fitting using summary descriptions 131

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

7.2 Parameter estimation using the K-function . . . . . . . . . . 132
7.2.1 Least squares estimation . . . . . . . . . . . . . . . . . 132
7.2.2 Simulated realisations of a Poisson cluster process . . 133
7.2.3 Procedure when K(t) is unknown . . . . . . . . . . . . 134
7.3 Goodness-of-fit assessment using nearest neighbour distribu-
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.1 Redwood seedlings . . . . . . . . . . . . . . . . . . . . 136
7.4.2 Bramble canes . . . . . . . . . . . . . . . . . . . . . . 139
7.5 Parameter estimation via goodness-of-fit testing . . . . . . . 147
xii Contents

7.5.1 Analysis of hamster tumour data . . . . . . . . . . . . 148

8 Model-fitting using likelihood-based methods 151

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

8.2 Likelihood inference for inhomogeneous Poisson processes . . 152
8.2.1 Fitting a trend surface to the Lansing Woods data . . 153
8.3 Likelihood inference for Markov point processes . . . . . . . 155
8.3.1 Maximum pseudo-likelihood estimation . . . . . . . . 156
8.3.2 Non-parametric estimation of a pairwise interaction
function . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.3.3 Fitting a pairwise interaction point process to the dis-
placed amacrine cells . . . . . . . . . . . . . . . . . . . 158
8.3.4 Monte Carlo maximum likelihood estimation . . . . . 160
8.3.5 The displaced amacrine cells re-visited . . . . . . . . . 163
8.3.6 A bivariate model for the displaced amacrine cells . . 165
8.4 Likelihood inference for Cox processes . . . . . . . . . . . . . 167
8.4.1 Predictive inference in a log-Gaussian Cox process . . 169
8.4.2 Non-parametric estimation of an intensity surface: hick-
ories in Lansing Woods . . . . . . . . . . . . . . . . . 170
8.5 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . 172

9 Point process methods in spatial epidemiology 173

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

9.2 Spatial clustering . . . . . . . . . . . . . . . . . . . . . . . . 176
9.2.1 Analysis of the North Humberside childhood leukaemia
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
9.2.2 Other tests of spatial clustering . . . . . . . . . . . . . 178
9.3 Spatial variation in risk . . . . . . . . . . . . . . . . . . . . . 179
9.3.1 Primary biliary cirrhosis in the North East of England 181
9.4 Point source models . . . . . . . . . . . . . . . . . . . . . . . 182
9.4.1 Childhood asthma in north Derbyshire, England . . . 185
9.4.2 Cancers in North Liverpool . . . . . . . . . . . . . . . 186
9.5 Stratification and matching . . . . . . . . . . . . . . . . . . . 189
9.5.1 Stratified case-control designs . . . . . . . . . . . . . . 189
9.5.2 Individually matched case-control designs . . . . . . . 191
9.5.3 Is stratification or matching helpful? . . . . . . . . . . 193
9.6 Disentangling heterogeneity and clustering . . . . . . . . . . 193

10 Spatio-temporal point processes 195

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

10.2 Motivating examples . . . . . . . . . . . . . . . . . . . . . . 196
10.2.1 Gastro-intestinal illness in Hampshire, UK . . . . . . . 196
Contents xiii

10.2.2 The 2001 foot-and-mouth epidemic in Cumbria, UK . 198

10.2.3 Bovine tuberculosis in Cornwall, UK . . . . . . . . . . 198
10.3 A classification of spatio-temporal point patterns and processes 199
10.4 Second-order properties . . . . . . . . . . . . . . . . . . . . . 202
10.5 Conditioning on the past . . . . . . . . . . . . . . . . . . . . 204
10.6 Empirical and mechanistic models . . . . . . . . . . . . . . . 207

11 Exploratory analysis 209

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

11.2 Animation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
11.3 Marginal and conditional summaries . . . . . . . . . . . . . . 210
11.3.1 Bovine tuberculosis in Cornwall, UK . . . . . . . . . . 210
11.4 Second-order properties . . . . . . . . . . . . . . . . . . . . . 213
11.4.1 Stationary processes . . . . . . . . . . . . . . . . . . . 213
11.4.2 Intensity-reweighted stationary processes . . . . . . . 216
11.4.3 Campylobacteriosis in Lancashire, UK . . . . . . . . . 217

12 Empirical models and methods 223

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

12.2 Poisson processes . . . . . . . . . . . . . . . . . . . . . . . . 224
12.3 Cox processes . . . . . . . . . . . . . . . . . . . . . . . . . . 224
12.3.1 Separable and non-separable models . . . . . . . . . . 225
12.4 Log-Gaussian Cox processes . . . . . . . . . . . . . . . . . . 226
12.5 Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
12.6 Gastro-intestinal illness in Hampshire, UK . . . . . . . . . . 227
12.7 Concluding remarks: point processes and geostatistics . . . . 230

13 Mechanistic models and methods 235

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

13.2 Conditional intensity and likelihood . . . . . . . . . . . . . . 235
13.3 Partial likelihood . . . . . . . . . . . . . . . . . . . . . . . . 237
13.4 The 2001 foot-and-mouth epidemic in Cumbria, UK . . . . . 238
13.5 Nesting patterns of Arctic terns . . . . . . . . . . . . . . . . 240

References 245

Index 263
This page intentionally left blank
List of Figures

1.1 Locations of 65 Japanese black pine saplings in a square of

side-length 5.7 metres (Numata, 1961). . . . . . . . . . . . . 2
1.2 Locations of 62 redwood seedlings in a square of side-length
23 metres (Strauss, 1975; Ripley, 1977). . . . . . . . . . . . . 3
1.3 Locations of 42 cell centres in a unit square (Ripley, 1977). . 4
1.4 Locations of 294 displaced amacrine cells in the retina of a
rabbit. Solid and open circles respectively identify on and off
cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Locations of cases of non-specific gastrointestinal symptoms
reported to NHS Direct Hampshire, UK, between 1 January
and 8 January 2001. The radius of each plotted circle codi-
fies the reporting date (smallest for 1 January, largest for 8
January). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Realisation of CSR: 100 events in a unit square. . . . . . . . 10
1.7 The Dirichlet tessellation (——) and Delaunay triangulation
(– – –) associated with 12 points in a unit square. . . . . . . 13
1.8 Blurred critical regions for one-sided, 5% Monte Carlo tests
with s = 20, 40, 100 and s → ∞ (adapted from Marriott,
1979). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.1 EDF plot of inter-event distances for Japanese black pine

saplings. —— : data; − − − : upper and lower envelopes from
99 simulations of CSR. . . . . . . . . . . . . . . . . . . . . . 20
2.2 EDF plot of inter-event distances for redwood seedlings. ——
: data; − − − : upper and lower envelopes from 99 simulations
of CSR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 EDF plot of inter-event distances for biological cells. —— :
data; − − − : upper and lower envelopes from 99 simulations
of CSR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 EDF plot of nearest neighbour distances for Japanese black
pine saplings. —— : data; − − − : upper and lower envelopes
from 99 simulations of CSR. . . . . . . . . . . . . . . . . . . 26
2.5 EDF plot of nearest neighbour distances for redwood
seedlings. —— : data; − − − : upper and lower envelopes
from 99 simulations of CSR. . . . . . . . . . . . . . . . . . . 27

xv
xvi List of Figures

2.6 EDF plot of nearest neighbour distances for biological cells.

—— : data; − − − : upper and lower envelopes from 99 sim-
ulations of CSR. . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.7 Calculation of F̂ (x) for the biological cells data, using a k × k
grid of sample points for different values of k. —— : k = 7;
− − − : k = 14; ....... : k = 96. . . . . . . . . . . . . . . . . . 29
2.8 EDF plot of point to nearest event distances for Japanese
black pine saplings. —— : data; − − − : upper and lower
envelopes from 99 simulations of CSR. . . . . . . . . . . . . 30
2.9 EDF plot of point to nearest event distances for redwood
seedlings. —— : data; − − − : upper and lower envelopes
from 99 simulations of CSR. . . . . . . . . . . . . . . . . . . 31
2.10 EDF plot of point to nearest event distances for biological
cells. —— : data; − − − : upper and lower envelopes from 99
simulations of CSR. . . . . . . . . . . . . . . . . . . . . . . . 32
2.11 Locations of trees in Lansing woods: left panel, hickories; mid-
dle panel, maples; right panel, oaks. . . . . . . . . . . . . . . 34
2.12 Index of dispersion plotted against block size (k × k) for Lans-
ing Woods data, based on a 32 × 32 grid of quadrat counts.
—— : hickories; − − − : maples; ....... : oaks. . . . . . . . . 36

3.1 The nearest event, P, to an arbitrary point, O, and the nearest

event, Q, to the event P. . . . . . . . . . . . . . . . . . . . . 45
3.2 T-square sampling: the nearest event, P, to an arbitrary point,
O, and the T-square nearest neighbour, Q, to the event P. . 46
3.3 Catana’s wandering quarter: solid lines indicate a sequence of
recorded distances, starting from the point O. . . . . . . . . 52

4.1 Construction of the edge-correction weights in Ripley’s (1976)

estimator for K(t) . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Comparison between Ripley’s asymptotic approximation and
the Lotwick-Silverman formula for the sampling variance of
K̂(t). —— : n = 100; − − − : n = 200; ....... : n = 400. The
horizontal line at height 1 corresponds to equality of the two
formulae. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3 Comparison between Chetwynd and Diggle’s estimator and
the Lotwick-Silverman formula for the sampling variance of
K̂(t). —— : n = 100; − − − : n = 200; ....... : n = 400. . . . 69
4.4 Comparison between the variances of K̃(t) and of K̂(t). ——
: n = 100; ......... : n = 200; – – – : n = 400 . . . . . . . . . . 70
4.5 Comparison between histogram-like (solid line) and kernel
smoothed estimates of the pair correlation function. Dashed
and dotted lines correspond to kernel estimates using a quartic
kernel and band-widths h = 0.038 and 0.076, respectively. . 72
List of Figures xvii

4.6 The estimate K̂(t) − πt2 for the Japanese black pine data. —
— : data; − − − : plus and minus two standard errors under
complete spatial randomness. . . . . . . . . . . . . . . . . . 74
4.7 Transformed estimates of K̂(t) for the Japanese black pine
data. —— : data; − − − : plus and minus two standard er-
rors under
√ complete spatial randomness. The left-hand panel
shows K̂(t), the right-hand panel {K̂(t) − πt2 }/t. . . . . . 75
4.8 The estimate D̂(t) = K̂(t) − πt2 for the redwood data (left-
hand panel) and for the cell data (right-hand panel). ——
: data; − − − : plus and minus two standard errors under
complete spatial randomness. . . . . . . . . . . . . . . . . . 76
4.9 Second-order properties of displaced amacrine cells. Functions
plotted are D̂(t) = K̂(t) − πt2 as follows: – – – : on cells; .......
: off cells; — — — : all cells; —— : bivariate. The parabola
−πt2 is also shown as a solid line. . . . . . . . . . . . . . . . 77
4.10 Estimates of F (x) (left-hand panel) and of G(x) (right-hand
panel) for the Japanese black pine data. —— : Ĝ(·), F̂ (·); – –
– : G̃(·), F̃ (·) . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.11 Estimates of F (x) (left-hand panel) and of G(x) (right-hand
panel) for the redwood data. —— : Ĝ(·), F̂ (·); – – – : G̃(·), F̃ (·) 81
4.12 Estimates of F (x) (left-hand panel) and of G(x) (right-hand
panel) for the cell data. —— : Ĝ(·), F̂ (·); – – – : G̃(·), F̃ (·) . 81

5.1 Estimates of the mean square error, M̂ (h) of a non-parametric

intensity estimator applied to each of the three major species
groupings in Lansing Woods. − − − − −− : hickories; − − −:
maples; ....... : oaks. . . . . . . . . . . . . . . . . . . . . . . . 87
5.2 Estimates K̂(t)−πt2 for each of the three major species group-
ings in Lansing Woods. —– : hickories; − − −: maples; .......
: oaks. Pointwise values of two standard errors assuming an
underlying homogeneous Poisson process are shown as − · − · − 88
5.3 Kernel estimates of λ(x) for the hickories (left-hand panel) and
maples (right-hand panel) in Lansing Woods, using a quartic
kernel with bandwidth h = 0.1. . . . . . . . . . . . . . . . . 89
5.4 Estimates of KI (t) for the hickories (left-hand panel) and
maples (right-hand panel) in Lansing Woods. − − − − − :
data; − − − upper and lower envelopes from 19 simulations
of the inhomogeneous Poisson process with intensity equal to
the corresponding kernel estimate λ̂(·). . . . . . . . . . . . . 91
5.5 Locations of pyramidal neurons in brain tissue sections from
12 normal subjects. . . . . . . . . . . . . . . . . . . . . . . . 92
5.6 Pooled estimate K̂(t) − πt2 from 12 control subjects, with
bootstrapped plus and minus two pointwise standard error
limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
xviii List of Figures

5.7 Locations of pyramidal neurons in brain tissue sections from

9 schizo-affective subjects. . . . . . . . . . . . . . . . . . . . 95
5.8 Locations of pyramidal neurons in brain tissue sections from
10 schizophrenic subjects. . . . . . . . . . . . . . . . . . . . 95
5.9 Pooled estimates K̂(t) − πt2 from 9 schizo-affective subjects
(left-hand panel) and from 9 schizophrenic subjects (right-
hand panel), with bootstrapped plus and minus two pointwise
standard error limits. . . . . . . . . . . . . . . . . . . . . . . 96

6.1 Realisations of two Poisson cluster processes, each with 25 par-

ents on the unit square an average of four offspring per parent
and radially symmetric Gaussian dispersion of offspring, with
parameter σ = 0.025. In the left-hand panel, each parent has
exactly four offspring. In the right-hand panel, offspring are
randomly allocated amongst the 25 parents. . . . . . . . . . 104
6.2 A realisation of an inhomogeneous Poisson process with 100
events on the unit square and λ(x) = exp(−2x1 − x2 ) . . . . 105
6.3 A realisation of a Cox process; see text for detailed explana-
tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.4 A realisation of a log-Gaussian Cox process (dots) superim-
posed on the realisation of the underlying intensity surface
(grey-scale image); see text for detailed explanation. . . . . 110
6.5 Progressive development of a realisation of a simple sequential
inhibition process with δ = 0.08 and n = 25, 50, 75, 100 events
on the unit square. . . . . . . . . . . . . . . . . . . . . . . . 112
6.6 Realisations of two pairwise interaction point processes on the
unit square and their corresponding interaction functions (see
text for detailed explanation). . . . . . . . . . . . . . . . . . 116
6.7 Realisations of two partly inhibitory, partly attractive pair-
wise interaction point processes on the unit square and their
interaction functions (see text for detailed explanation). . . 117
6.8 A thinning of an inhibitory pairwise interaction point process
(see text for detailed explanation). . . . . . . . . . . . . . . 120
6.9 Realisations of two processes with identical second-order prop-
erties (upper panels) and their estimated nearest neighbour
distributions functions (lower panels), with dashed and solid
lines corresponding to a Poisson cluster process and the su-
perposition of a Poisson cluster process and a homogeneous
Poisson process. . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.10 An inhomogeneous pairwise interaction point process (see text
for detailed explanation). . . . . . . . . . . . . . . . . . . . . 123
6.11 A linked (left-hand panel) and a balanced (right-hand panel)
log-Gaussian Cox process (see text for detailed explanation). 127
List of Figures xix

6.12 Two mutually inhibitory bivariate pairwise interaction point

processes. The stronger inhibition is between events of oppo-
site type in the left-hand panel, and between events of the
same type in the right-hand panel (see text for detailed ex-
planation). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

7.1 The empirical sampling distribution of (log ρ̂, log σ̂) in a Pois-
son cluster process with 400 events in the unit square, ρ = 100
and σ = 0.025. . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.2 K(t) − πt2 for the redwood seedlings. —–: data; − − −: fitted
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.3 Goodness-of-fit of a Poisson cluster process to the redwood
seedling data, using nearest neighbour (left-hand panel) and
point-to-nearest-event (right-hand panel) distribution func-
tions. —— : data; − − − : envelope from 99 simulations of
fitted model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.4 A realisation of the model fitted to the redwood seedling data. 139
7.5 The empirical sampling distribution of (log ρ̂, log σ̂) for the
model fitted to the redwood seedling data. . . . . . . . . . . 140
7.6 Locations of 359 newly emergent bramble canes in a 9 metre
square plot (Hutchings, 1978) . . . . . . . . . . . . . . . . . 141
7.7 K(t)−πt2 for newly emergent bramble canes (solid black line),
fitted model (dashed black line) and 99 simulations of fitted
model cluster process (thin grey lines). In the left-hand panel,
the fitted model is a two-parameter Poisson cluster process;
in the right-hand panel, it is four-parameter thinned Poisson
cluster-process. . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.8 Goodness-of-fit for the four-parameter model fitted to the
newly emergent bramble canes, using nearest neighbour (left-
hand panel) and point-to-nearest-event (right-hand panel) dis-
tribution functions. The data are shown as thick black lines,
simulations of the fitted model as thin grey lines. . . . . . . 143
7.9 A realisation of the four-parameter model fitted to the newly
emergent bramble canes . . . . . . . . . . . . . . . . . . . . 144
7.10 Locations of 359 newly emergent (solid dots) and 385 one-
year-old (open circles) bramble canes in a 9 metre square plot
(Hutchings, 1978). . . . . . . . . . . . . . . . . . . . . . . . 145
7.11 K̂ij (t) − πt2 for newly emergent and one-year-old bramble
canes: solid line corresponds to K11 (t) (newly emergent),
dashed line to K22 (t) (one-year-old), dotted line to K12 (t). . 146
xx List of Figures

7.12 Left-hand panel: K̂12 (t) − πt2 for newly emergent and one-
year-old bramble canes (thick black line) and for simulations
of fitted linked Cox process model (thin grey lines). The fit-
ted function K12 (t) − πt2 is shown as a thick dashed line.
Right–hand panel: Ĥ2 (x) for newly emergent and one-year-
old bramble canes (solid black line) and for 99 simulations of
fitted linked Cox process model (thin grey lines). . . . . . . 147
7.13 Ĥ3 (x) for newly emergent, one-year-old and two-year-old
bramble canes (solid black line) and for 99 simulations (thin
grey lines). In the left-hand panel, the simulation model is the
fitted trivariate linked Cox process model. In the centre and
right-hand panels the simulation model is random labelling. 148
7.14 Locations of 303 cell nuclei in a hamster tumour; 77 pyknotic
nuclei (solid dots); 226 metaphase nuclei (open circles) . . . 149
7.15 Dij (t) = K̂ij (t) − πt2 for the hamster tumour data, com-
pared with 99 simulations of a pair of independent homo-
geneous Poisson processes. Upper-left panel shows K̂11 (py-
knotic cells), upper-right panel K̂22 (metaphase cells), lower-
left panel K̂12 , lower-right panel K̂ (superposition). . . . . . 150
7.16 Left-hand panel: 95% confidence interval for interaction pa-
rameter θ in pairwise interaction process with h(u) = 1 −
exp{−(u/θ)2 }, fitted to the hamster tumour data. Right-
hand panel: goodness-of-fit for hamster tumour data, using
the nearest neighbour distribution function. Thick black line
is calculated from the data, thin grey lines from 99 simulations
of pairwise interaction process with θ = 0.014. . . . . . . . . 150

8.1 Fitted log-quadratic trend surfaces for the hickories (left-hand

panel) and maples (right-hand panel) in Lansing Woods . . 154
8.2 Estimates of KI (t) for the hickories (left-hand panel) and
maples (right-hand panel) in Lansing Woods. Data are shown
as thick black lines, simulations of the fitted log-quadratic in-
homogeneous Poisson process as thin grey lines. . . . . . . . 155
8.3 Non-parametric (step-function) estimate of the interaction
function for the displaced amacrine on cells. . . . . . . . . . 159
8.4 Parametric (dashed line) and non-parametric (step-function)
estimates of the interaction function for the displaced
amacrine on cells. . . . . . . . . . . . . . . . . . . . . . . . . 160
8.5 Profile log-pseudo-likelihood surface for the pairwise interac-
tion point process fitted to the on cells. The plotted value at
each point is the log-pseudo-likelihood maximised with respect
to β, holding δ and ρ fixed. . . . . . . . . . . . . . . . . . . 161
List of Figures xxi

8.6 Goodness-of-fit assessment for displaced amacrine off cells, us-

ing point-to-nearest-event (left-hand panel) and nearest neigh-
bour (right-hand panel) distribution functions. —— : data;
− − −: envelope from 99 simulations of fitted pairwise inter-
action point process. . . . . . . . . . . . . . . . . . . . . . . 162
8.7 Monte Carlo log-likelihood functions for the displaced
amacrine on cells, with interaction function (8.10), θ0 =
(δ, ρ, β) = (0.016, 0.121, 2.091) and s = 10 simulations per
log-likelihood evaluation. Thin grey lines show each of five in-
dependent replicates, thick black line the point-wise average
of the five. Left-hand panel: L̂θ0 ,10 (δ). Centre panel: L̂θ0 ,10 (ρ).
Right-hand panel: L̂θ0 ,10 (β). . . . . . . . . . . . . . . . . . . 163
8.8 Monte Carlo log-likelihood functions for the displaced
amacrine on cells, with interaction function (8.10), θ0 =
(δ, ρ, β) = (0.017, 0.122, 2.100) and s = 100 simulations per
log-likelihood evaluation. Thin grey lines show each of five in-
dependent replicates, thick black line the point-wise average of
the five. Left-hand panel: L̂θ0 ,100 (δ). Centre panel: L̂θ0 ,100 (ρ).
Right-hand panel: L̂θ0 ,100 (β). . . . . . . . . . . . . . . . . . 164
8.9 Monte Carlo log-likelihood functions for the displaced
amacrine on cells, with interaction function (8.10), θ0 =
(δ, ρ, β) = (0.017, 0.122, 2.100) and s = 1000 simulations per
log-likelihood evaluation. Thin grey lines show each of five
independent replicates, thick black line the point-wise aver-
age of the five. Left-hand panel: L̂θ0 ,1000 (δ). Centre panel:
L̂θ0 ,1000 (ρ). Right-hand panel: L̂θ0 ,1000 (β). . . . . . . . . . . 165
8.10 Schematic representation of the placement of displaced
amacrine cells in the transition from the immature to the ma-
ture retina. . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.11 Prior (smooth curves) and posterior (histograms) distribu-
tions for the parameters β, σ and φ in the LGCP model for
the hickory data. . . . . . . . . . . . . . . . . . . . . . . . . 171
8.12 Predictive probabilities P{exp[S(x)] < 1/2|data} (left-hand
panel), pointwise medians of the predictive distribution
of exp{β + S(x)} (centre panel), predictive probabilities
P{exp[S(x)] > 2|data} (right panel). Left and right panels
also show data. . . . . . . . . . . . . . . . . . . . . . . . . . 172

9.1 Residential locations at birth of 62 cases of childhood

leukaemia in North Humberside, UK (left-hand panel), and
of 143 controls sampled at random from the birth register
(right-hand panel). . . . . . . . . . . . . . . . . . . . . . . . 175
xxii List of Figures

9.2 Second-order analysis of clustering for the North Humberside

childhood leukaemia data. —— : D̂(t) for observed data; −−−
: plus and minus two standard errors under random labelling
of cases and controls. . . . . . . . . . . . . . . . . . . . . . . 177
9.3 Residential locations (postcodes) of 761 cases of primary bil-
iary cirrhosis in North East England (solid dots) and 3044
control locations selected by weighted random sampling from
the postcode directory (open circles). The dashed line in the
left-hand panel indicates the approximation to the boundary
of the study-region used in the analysis of the data. The right-
hand panel shows the densely populated part of the study-
region in more detail. . . . . . . . . . . . . . . . . . . . . . . 181
9.4 Cross-validated log-likelihood criterion for choice of band-
width in non-parametric estimation of the relative risk surface
for primary biliary cirrhosis in the North East of England. . 182
9.5 Estimated relative risk surface for primary biliary cirrhosis in
the North East of England. The scale denotes the probability,
at each location x in the study-region, that a case or control
at location x is a case. . . . . . . . . . . . . . . . . . . . . . 183
9.6 Examples of a directional model for elevation in risk around
a point source. . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.7 Study region, control locations (solid dots) and incinerator
location (solid triangle) for the North Liverpool cancer study. 188
9.8 Estimated residual spatial variation in risk for colorectal can-
cers in the North Liverpool cancer study. The risk scale is
logarithmic to base 2. Solid and dashed contours identify re-
gions within which risk is point-wise 5% significantly higher or
lower, respectively, than the average for the whole study-area. 189
9.9 Estimated residual spatial variation in risk for lung cancers in
the North Liverpool cancer study. The risk scale is logarithmic
to base 2. Solid and dashed contours identify regions within
which risk is point-wise 5% significantly higher or lower, re-
spectively, than the average for the whole study-area. . . . . 190
9.10 Estimated residual spatial variation in risk for leukaemias/
lymphomas in the North Liverpool cancer study. The risk scale
is logarithmic to base 2. Solid and dashed contours identify
regions within which risk is point-wise 5% significantly higher
or lower, respectively, than the average for the whole study-
area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

10.1 Locations (residential post-codes) of 10,572 successive cases

of non-specific gastro-intestinal disease, as reported to NHS
Direct in Hampshire, UK, between 1 January 2001 and 31
December 2003. . . . . . . . . . . . . . . . . . . . . . . . . . 196
List of Figures xxiii

10.2 Times (days since 1 January 2001) of 10,572 successive cases

of non-specific gastro-intestinal disease, as reported to NHS
Direct in Hampshire, UK, between 1 January 2001 and 31
December 2003. . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.3 Data from the 2001 UK foot-and-mouth epidemic. Left panel:
small dots show the locations of all animal-holding farms in
the county of Cumbria, larger dots show the locations of those
farms that experienced foot-and-mouth during the epidemic.
Right panel: cumulative distribution of reporting times of
cases, in days since 1 February 2001. . . . . . . . . . . . . . 199
10.4 Locations of farms in Cornwall, UK, that tested positive for
bovine tuberculosis at least once during the period 1989 to
2002. The grey-scale shading of the locations corresponds to
the year in which each farm first tested positive (from white
in 1989 to black in 2002). . . . . . . . . . . . . . . . . . . . 200
10.5 Number of farms in Cornwall, UK, testing positive for bovine
tuberculosis in each year, 1989 to 2002. . . . . . . . . . . . . 201
10.6 Snap-shots from a simulated partial realisation of a continuous
spatio-temporal pairwise interaction point process with both
inhibitory and aggregated characteristics. In each frame, the
most recent 30 or 40 events are shown as solid dots, earlier
events as open circles. See text for details of the simulation
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
10.7 Snap-shots from a simulated partial realisation of a continuous
spatio-temporal pairwise interaction point process with purely
inhibitory characteristics. In each frame, the most recent 30
or 40 events are shown as solid dots, earlier events as open
circles. See text for details of the simulation model. . . . . . 207

11.1 Annual numbers of incident bovine tuberculosis cases in Corn-

wall, relative to the average annual incidence over the baseline
period 1989 to 1995. . . . . . . . . . . . . . . . . . . . . . . 211
11.2 Cross-validated log-likelihood criteria, LCV (h), for kernel
smoothing over bovine tuberculosis cases in Cornwall. Thin
lines refer to comparisons between baseline period 1989 to
1995 and individual post-baseline years, whilst the thick line
is calculated as the average of the LCV (h) over all seven post-
baseline years. Dotted lines identify the location and value of
the maximised average. . . . . . . . . . . . . . . . . . . . . . 212
11.3 Kernel smoothing estimates of annual incidence of bovine tu-
berculosis cases in Cornwall, 1996 to 1999. In each case, the
comparison is between the year indicated and the seven-year
baseline period 1999 to 1995 and the common grey-scale runs
from 0 (black) to 1 (white). The value 1/7 ≈ 0.14 corresponds
to equality between current and baseline incidence. . . . . . 213
xxiv List of Figures

11.4 Kernel smoothing estimates of annual incidence of bovine tu-

berculosis cases in Cornwall, 2000 to 2002. In each case, the
comparison is between the year indicated and the seven-year
baseline period 1999 to 1995 and the common grey-scale runs
from 0 (black) to 1 (white). The value 1/7 ≈ 0.14 corresponds
to equality between current and baseline incidence. . . . . . 214
11.5 Campylobacteriosis in the district of Preston, 2000-2002. Lo-
cations of cases (left panel) cumulative distribution of report-
ing times in days since 1 January 2000 (right panel). . . . . 218
11.6 Lancashire campylobacteriosis data: study-area (left panel);
2001 population density in number of people per hectare (cen-
tre panel); locations of the 619 cases within the urban area
(right panel). . . . . . . . . . . . . . . . . . . . . . . . . . . 218
11.7 Lancashire campylobacteriosis data: kernel estimate of spatial
intensity (left panel); weekly numbers of notified cases and
fitted regression curve (right panel). . . . . . . . . . . . . . . 219
11.8 Lancashire campylobacteriosis data: K̂st (u, v) − πu2 v, ×106
(left panel); identification of sub-sets of (u, v)-space in which
spatio-temporal clustering (diagonal black hatching) and/or
spatio-temporal interaction (grey shading) is detected at the
5% level of significance (right panel). . . . . . . . . . . . . . 220

12.1 Gastro-intestinal illness in Hampshire: observed counts of re-

ported cases per day, averaged over successive weekly peri-
ods (solid dots), compared with the fitted harmonic regression
model of daily incidence (solid line). . . . . . . . . . . . . . 229
12.2 Gastro-intestinal illness in Hampshire: adaptive kernel esti-
mate of the overall spatial variation in incidence, ρ̂(x). . . . 230
12.3 Gastro-intestinal illness in Hampshire: non-parametric (solid
line) and fitted parametric (dashed line) log-transformed pair
correlation functions (left-hand panel) and temporal covari-
ance functions (right-hand panel). . . . . . . . . . . . . . . . 231
12.4 Gastro-intestinal illness in Hampshire: predictive map for 6
March 2003. The mapped value in each case is the conditional
probability that R(x, t) > 2 given the data up to and including
day t. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
12.5 Three simulated transformed Gaussian processes, R(x, t) =
ak + bk S(x, t)2k , for k = 0.5, 1, 2 (solid, dashed and dotted
lines, respectively). . . . . . . . . . . . . . . . . . . . . . . . 233

13.1 Estimated transmission kernel in the spatial SIR model fitted

to data on the 2001 FMD epidemic in Cumbria, UK. . . . . 240
13.2 Locations of Arctic tern nests (crosses) on each of two islets in
the Abro Delta National Park, Spain, and additional locations
at which height above sea-level was also recorded (triangles). 241
List of Figures xxv

13.3 Estimated interaction function, h(d), for the model fitted to

the Arctic tern nesting data (thick line) and re-estimates from
100 simulated realisations of the model (thin grey lines). . . 243
This page intentionally left blank
List of Tables

2.1 Analysis of Lansing Woods data by Mead’s procedure, using

a 32 × 32 grid of quadrat counts and 99 randomizations. A
high rank for u1 suggests aggregation. Where ties occur, the
least extreme rank is quoted. Block size refers to the number
of quadrats which are treated as a single 4 × 4 block. . . . . 35

3.1 Quadrat count analysis of Lansing Woods data, using 25

square quadrats of side 0.05. . . . . . . . . . . . . . . . . . . 43
3.2 T-square analysis of Lansing Woods data. Figures in paren-
theses indicate attained significance levels for tests of CSR
(two-sided for tN , one-sided for M ). . . . . . . . . . . . . . . 51

7.1 Estimated correlations amongst attained significance levels for

three goodness-of-fit tests . . . . . . . . . . . . . . . . . . . 136
7.2 Empirical standard errors of, and correlations amongst, pa-
rameter estimates for model fitted to newly emergent bramble
canes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

8.1 Maximised log-likelihoods for Poisson log-quadratic trend sur-

face models fitted to the Lansing Woods data . . . . . . . . 154

9.1 Deviances (twice maximised log-likelihoods) for various sub-

models fitted to the north Derbyshire childhood asthma data 186
9.2 Numbers of North Liverpool cancer cases available for analysis 187
9.3 Parameter estimates of regression effects for age, sex and
Townsend deprivation index (TI) in the North Liverpool can-
cer study, and p-values for associated likelihood ratio tests of
significance. . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

13.1 Parameter estimation for the five-parameter model fitted to

combined data from Cumbria and Devon . . . . . . . . . . . 240
13.2 Maximum partial likelihood estimates and standard errors for
the model fitted to the Arctic tern nesting data. Standard er-
rors are Monte Carlo approximations, calculated by re-fitting
the model to 100 simulated data-sets. . . . . . . . . . . . . . 243

xxvii
This page intentionally left blank
Preface

A spatial point pattern is a set of locations, irregularly distributed within

a designated region and presumed to have been generated by some form of
stochastic mechanism. In most applications, the designated region is essen-
tially planar (two-dimensional Euclidean space), but one-dimensional appli-
cations are also possible, and three-dimensional applications are becoming
more common in conjunction with the development of more sophisticated
three-dimensional scanning microscopes. The first edition of Statistical Anal-
ysis of Spatial Point Patterns appeared in 1983. Its aim was to cover the major
methodological themes of the subject and their application to data arising in
the biological sciences, especially ecology.
In the second edition, published in 2003, I extended the methodological
discussion to cover major developments in the intervening years, but also tried
to preserve the applied flavour of the book. Much of the newer work in the
area tends to be mathematically sophisticated. My aim in covering the newer
methodological developments was to discuss the central ideas without going
into the full technical detail which a rigorous treatment would require, and
which is available in the original articles. I also resisted the temptation to
discuss spatial statistics more widely. Cressie (1991) identified the three main
branches of spatial statistics as geostatistics (spatially continuous processes),
lattice processes (spatially discrete processes) and spatial point processes.
Whilst these three topics are to some extent inter-linked, they nevertheless
give rise to distinct stochastic models and associated statistical methods, and
can therefore be studied separately.
Within the realm of spatial point processes, perhaps the most important
theoretical development over the last three decades has been the provision
of formal, likelihood-based methods of inference for a reasonably wide range
of models. These have partially replaced the more ad hoc methods which
prevailed in the early 1980’s. Nevertheless, some of the ad hoc methods remain
useful, and have themselves been extended in various ways, for example in the
adaptation of non-parametric smoothing methods to spatial point processes.
New applications have also emerged and, as is usual in statistics, have in
turn motivated further methodological development. The two new areas of
application on which I draw most heavily, in my own research and in the
book, are micro anatomy and epidemiology.
In microanatomy, the points in an observed pattern typically are refer-
ence locations for cells in a microscopic tissue section. The underlying cellular
structure influences the kinds of models which are appropriate, usually in-

xxix
xxx Preface

volving the concept of interactions between near-neighbouring cells. Perhaps

more fundamentally from a statistical viewpoint, most micro-anatomical stud-
ies use a replicated sampling design in which data are obtained from several
subjects and/or several tissue sections per subject, in contrast to the tradi-
tional emphasis throughout spatial statistics on the analysis of unreplicated
patterns.
In epidemiology, the points are reference locations (typically place of res-
idence) for cases of a disease in a geographical region, often supplemented
by the reference locations for a set of controls sampled from the underlying
population at risk. The principal methodological challenge in this area of appli-
cation is to use the case-control paradigm to circumvent the obvious difficulty
of building credible parametric models for human population distributions in
a heterogeneous environment.
When the first edition was written, there were few other books available
on the then-emerging subject of spatial statistics, and none at all which dealt
exclusively with the statistical analysis of spatial point patterns. This is no
longer the case. Books on spatial point processes and associated statistical
methods include Møller and Waagepetersen (2004) and Ilian et al (2008),
whilst Van Lieshout (2000) deals exclusively with Markov point processes
and their statistical analysis. The first general text on spatial statistics was
Ripley (1981). This, and Cressie (1991), remain standard references. Other
books on spatial statistics that include substantial material on point processes
include the two volumes of Upton and Fingleton (1985, 1989), Cliff and Ord
(1981), Bailey and Gatrell (1995), Stoyan, Kendall and Mecke (1995), Waller
and Gotway (2004), Schabenberger and Gotway (2004), and Gelfand et al.
(2010). Additionally, and typically for a maturing field, a wide variety of more
specialised books have also become available. For example, Matérn (1986) is a
re-issue of Bertil Matérn’s classic 1960 Swedish doctoral dissertation that laid
many of the foundations for later developments in spatial point processes and
geostatistics. Rue and Held (2005) discuss Gaussian Markov random fields, the
most widely used modelling framework for spatially discrete processes. Chilès
and Delfiner (1999, 2012) and Diggle and Ribeiro (2007) cover the so-called
“classical” and “model-based” approaches to geostatistics, respectively.
Some topics which seemed important at the beginning of the 1980’s had
become less so by 2003, and I reduced the space given to them. One example
was the discussion of methods for sparse sampling of spatial point processes in
situ. These methods arose in the 1950’s and 1960’s, principally in connection
with field-work by ecologists, who needed to investigate the density and spatial
pattern of plant communities in the field. By 2003, these methods were rarely
used because technological advances had made much more straighforward the
task of mappping a spatial point pattern for later analysis using the more
sophisticated methods that were by then available. To my surprise, however,
these methods subsequently experienced a minor revival when they began to
be used for analysing the size and pattern of refugee encampments (see, for
example, Bostoen, Chalabi and Grais, 2007).
Preface xxxi

Thus forewarned, in this new book I have not deleted any material from
the second edition, other than to correct a number of errors, but I have added
substantial new material in places.
The biggest change from the second edition, reflected in the enlarged title,
is to discuss spatio-temporal point patterns. Spatio-temporal point process
data have long been studied in specialised fields, notably seismology (see, for
example, Zhuang, Ogata and Vere-Jones, 2002). However, in the last decade
there has been an acceleration of methodological development, accompanied
by a diversification of application as spatio-temporially indexed data have
become more widely available in many scientific fields. Book-length treatments
are now beginning to appear, including the edited collection by Finkenstadt,
Held and Isham (2007), several chapters of Gelfand et al. (2009) and, most
recently, Cressie and Wikle (2011).
Another important development, throughout the statistics discipline, has
been the rise in popularity of R as a vehicle for the dissemination of new statis-
tical methods through open-source software. Useful packages for the analysis of
spatial point process data include spatial, spatstat, MarkedPointProcess,
splancs and spatialkernel. All of these, and more, can be downloaded from
the R project web-page, www.r-project.org. I predict with some confidence
that the above list will be out-of-date by the time this appears in print.
Public-domain data-sets used in the book, and any errors of which I am
aware, can be found on the book’s web-page:
http://wwww.lancs.ac.uk/staff/diggle/pointpatternbook
My thanks are again due to many colleagues, in many places and over
some forty years, who have provided me with such stimulating working envi-
ronments, spanning the UK, Sweden, Australia and the USA. I was fortunate
to begin my career under the wise guidance of the late Prof Robin Plackett
at the University of Newcastle upon Tyne. Periods spent at the Royal College
of Forestry Stockholm, CSIRO Australia and, most recently, the University
of Liverpool, have taught me the inestimable value of working closely with
subject-matter scientists. Visits to the Department of Biostatistics at Johns
Hopkins University, Baltimore, stimulated an enduring interest in medical and
public health applications. At Lancaster University, I have been privileged to
work with a succession of talented young research students and staff, amongst
whom special mention goes to Barry Rowlingson for his patient, if doomed,
efforts over 25 years to teach me to compute efficiently.
Finally, my collaborators on the many jointly authored publications listed
amongst the references should share the credit for whatever value the book
may have, whereas responsibility for defects remains mine alone.

Peter J. Diggle, Lancaster

This page intentionally left blank
1
Introduction

CONTENTS
1.1 Spatial point patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Edge-effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Complete spatial randomness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Objectives of statistical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 The Dirichlet tessellation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Monte Carlo tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.1 Spatial point patterns

Data in the form of a set of points, irregularly distributed within a region of
space, arise in many different contexts; examples include locations of trees in
a forest, of nests in a breeding colony of birds, or of nuclei in a microscopic
section of tissue. We call any such data-set a spatial point pattern and refer
to the locations as events, to distinguish these from arbitrary points of the
region in question.
Figures 1.1 and 1.2 show two spatial point patterns in a square region.
The first, due to Numata (1961), shows 65 Japanese black pine saplings in a
square of side 5.7 metres whilst the second, extracted by Ripley (1977) from
Strauss (1975), shows 62 redwood seedlings in a square of side 23 metres, ap-
proximately. The two patterns appear strikingly different. Figure 1.1 shows no
obvious structure and might be regarded as a “completely random” pattern, in
a sense that we shall define formally in due course. In Figure 1.2, on the other
hand, the strong clustering of seedlings requires some biological explanation
which, in this instance, is readily available. The seedlings cluster around red-
wood stumps, which are known to be present in the study region but whose
locations have not been recorded. It is important to recognise that patterns
like Figure 1.2 can arise either through some form of clustering mechanism or
through environmental variation leading to local patches with relatively high
concentrations of events. Here, as elsewhere, failure to record relevant biolog-
ical information limits the conclusions which can be drawn from a statistical

1
2 Statistical Methods for Spatial and Spatio-Temporal Point Processes

• • •• • •
• • •
• ••• • •
• • ••
• •
•• • • • •
• •• •
• • •
•• • •
•
•• • • •
•
• • • •
• • •• •
• •
• •
• • • • • •

FIGURE 1.1
Locations of 65 Japanese black pine saplings in a square of side-length 5.7
metres (Numata, 1961).

analysis. We therefore describe patterns like Figure 1.2 as “aggregated” to

avoid the mechanistic connotations of the perhaps more obvious term “clus-
tered.”
Figure 1.3 shows a further qualitatively different type of pattern, formed
in this case by the centres of 42 biological cells (Crick and Lawrence, 1975;
Ripley, 1977). The cell centres are distributed more or less regularly over the
unit square, improbably so unless there is some associated regulating mech-
anism operating to encourage an even spatial distribution of cell centres. A
possible explanation is that the cell centres are merely convenient reference
points for cells whose physical size is non-negligible relative to the scale of
observation. Quite generally, and again without wishing to imply any specific
causal mechanism, we refer to such patterns as “regular.”
The nature of the pattern generated by a biological process can be affected
by the physical scale on which the process is observed. At a sufficiently large
scale most natural environments exhibit heterogeneity, which will tend to pro-
duce aggregated patterns. At a smaller scale, environmental variation will be
Introduction 3

•
• •
• •• • • • • • •
• ••
• ••
• ••
• •
••• • •• •
••
•• ••
•
•••
••• •
• ••
•
• ••
• •••
• • •• • ••

FIGURE 1.2
Locations of 62 redwood seedlings in a square of side-length 23 metres
(Strauss, 1975; Ripley, 1977).

less pronounced and the major determinant of pattern may then be the nature
of the interactions amongst the events themselves. For example, vegetative
propagation of individual shoots will tend to produce small-scale aggregation
whereas competition for space will encourage regularity. Our classification of
patterns as regular, random or aggregated is therefore an over-simplification,
but a useful one at an early stage of analysis. At a later stage, this simplistic
approach can be abandoned in favour of a more detailed, and essentially mul-
tidimensional, description of pattern that can be obtained either by the use of
a variety of functional summary statistics or by formulating an explicit model
of the underlying process. The approach taken in this book will be to develop
methods for the analysis of spatial patterns based on stochastic models, which
assume that the events are generated by some underlying random mechanism.
Our fourth example, shown in Figure 1.4, introduces the idea of a multivari-
ate point pattern. In this example, the points represent cells of two different
types (hence, bivariate) in the retina of a rabbit. The data consist of the lo-
cations of 294 displaced amacrine cells, amongst which 152 are of a type that
4 Statistical Methods for Spatial and Spatio-Temporal Point Processes

• •
• •
•
• •
• • • •
•
• • •
• •
• • • •
• •
• •
• • •
•
• • •
• •
• •
• •
•
• • •

FIGURE 1.3
Locations of 42 cell centres in a unit square (Ripley, 1977).

transmits information to the brain when a light goes on, whilst the remaining
142 transmit information when a light goes off. The relationship between the
two component patterns can help to explain the developmental processes that
operate within the immature retina. We shall re-examine the data from this
point of view in Section 4.7.
Our fifth example is of a spatio-temporal point pattern, in which the data
provide both the location and the time of occurrence of events of scientific
interest within a specified spatial region and time-interval. Figure 1.5 shows
the residential locations and dates of 100 consecutive cases of non-specific
gastrointestinal symptoms, as reported between 1 and 8 January 2001 to NHS
Direct, a 24-hour phone-based triage service operated by the UK National
Health Service, by residents in the county of Hampshire. The cases naturally
cluster in areas of relatively high population density, but there is at least a
hint that cases close in time (circles of the same radius of nearly so) are also
closer spatially than might be expected by chance. If true, this would suggest
that multiple cases may be the result of infections from a common source.
Spatio-temporal patterns are better examined dynamically than statically.
Introduction 5

• • •
• • • • • • • • • ••
•• • • • • • • • • • • • • • •
• • • • • • • •• •
• • •• • •
• • • • • • • • •• • •
• • • •• • • • ••
• • • • • • • •• • • • •
• • • • • •• • • • • • •
• • • • • • • •• •
•
• • • • • • •• •• • • •• •
• • • • • • • •••
• • • • •• • • • •
• • • • • • •• • • • •

FIGURE 1.4
Locations of 294 displaced amacrine cells in the retina of a rabbit. Solid and
open circles respectively identify on and off cells.

The data shown in Figure 1.5 are a sub-set of a much larger data-set reported
in Diggle et al. (2003); an animation of the complete data-set by Barry Rowl-
ingson can be viewed from the book’s web-site.
We shall assume throughout this book that the spatial region of interest is
essentially planar, although most of the ideas extend, at least in principle, to
other dimensions. Even in one dimension, the distinction between temporal
and spatial point patterns is important. In the case of series of events irreg-
ularly distributed in time, for example division times in a cell proliferation
process, stochastic models and their associated statistical methods reflect the
essentially unidirectional quality of the time dimension, whereas in the corre-
sponding spatial case, for example nesting sites along the bank of a canal, no
such directionality exists. Cox and Lewis (1966) give an excellent introduction
to the analysis of temporal point patterns, whilst Daley and Vere-Jones (2002,
2005) discuss the underlying point process theory in depth.
All of our examples involve applications in the life sciences, although sim-
ilar problems arise in many other disciplines. For examples in archaeology,
astronomy and geography see, respectively, Hodder and Orton (1976), Pee-
bles (1974) and Cliff and Ord (1981). To some extent, the methods that we
6 Statistical Methods for Spatial and Spatio-Temporal Point Processes

180000
160000
140000
N−S

120000
100000
80000

400000 420000 440000 460000 480000 500000

E−W

FIGURE 1.5
Locations of cases of non-specific gastrointestinal symptoms reported to NHS
Direct Hampshire, UK, between 1 January and 8 January 2001. The radius of
each plotted circle codifies the reporting date (smallest for 1 January, largest
for 8 January).

describe remain useful (and have certainly been used) in these other areas of
application, but should not be adopted uncritically. In particular, our stochas-
tic models will be motivated by simple considerations of possible underlying
biological mechanisms that may or may not be relevant in other disciplines.

1.2 Sampling
The selection of the study region, A say, merits some discussion. In some
applications, A is objectively determined by the problem in hand, and infer-
ences are required in terms of a process defined on A itself. One example of
Introduction 7

this would be a map of all nesting sites on an island. More commonly, A is

selected from some much larger region. The selection of A may then be made
according to a probability sampling scheme, or it may simply reflect the ex-
perimenter’s view that A is in some sense representative of the larger region.
In either case, but particularly the latter, inferences drawn from an analysis
will carry much greater conviction if consistency over replicate data-sets can
be demonstrated.
As an alternative to intensive mapping within a single region A, the ex-
perimenter may choose to record limited information from a large number of
smaller regions, for example the number of events in each region. In this con-
text, the regions are called quadrats and the data are referred to as quadrat
counts. A quadrat was originally a square of side-length 1 metre, used by the
Uppsala school of plant ecologists as the basic sampling unit for investigating
plant communities in the field (Du Rietz, 1929).
Quadrat sampling remains a popular field technique in plant ecology and
elsewhere, but in some contexts it is rather impractical. This led to the de-
velopment, initially in the American forestry literature (Cottam and Curtis,
1949), of a number of distance methods for sampling spatial point patterns.
In these, the basic sampling unit is a point, and information is recorded in
the form of distances to neighbouring events, for example the distances to the
first few nearest events.
We shall refer to quadrat count and distance methods as sparse sampling
methods, to distinguish them from intensive mapping exercises. The appropri-
ate techniques for analysing data obtained by sparse sampling and by intensive
mapping are quite different. Also, analyses of sparsely sampled patterns typ-
ically have more limited objectives than do analyses of mapped patterns.
A particular form of quadrat sampling, intermediate between sparse sam-
pling and intensive mapping, consists of partitioning the study region into dis-
joint sub-regions and recording the number of events in each sub-region. Data
of this kind arise in two very different ways. The first, which has a long tradi-
tion as a method of field sampling in plant ecology, is when a rectangular study
region is partitioned into a regular grid of square or rectangular quadrats,
and a count is taken within each quadrat. The second, which typically arises
in environmental epidemiology, is when health outcome data are routinely
maintained as counts of the numbers of events in administratively defined
sub-regions. In either case, the resulting data can be represented as a realisa-
tion of a high-dimensional multivariate random variable, Y = (Y1 , ..., Yn ) say,
where Yi denotes the number of events in the ith sub-region. In this setting, a
stochastic model for the underlying point process would induce a unique sta-
tistical distribution on Y , but in practice the form of this distribution tends
to be intractable except in a few special cases. A pragmatic alternative is
to formulate a model directly for the distribution of Y , without reference to
any underlying point process. The usual method of construction is to specify
the set of conditional distributions of each Yi given all other Yj . Models of
this kind are called Markov random fields. Their construction must satisfy
8 Statistical Methods for Spatial and Spatio-Temporal Point Processes

sometimes non-obvious constraints to ensure self-consistency. Besag (1974) is

an early, and very influential, account; see also Rue and Held (2005) or the
relevant chapters of Gelfand et al. (2009) for detailed accounts.
Replicated sampling of mapped patterns is surprisingly rare. Ecological
investigations have certainly compared patterns in study regions deliberately
selected to represent different environmental conditions (see, for example,
Bagchi et al. 2011), but I am not aware of corresponding studies which have
been designed with a view to establishing the consistency of patterns in os-
tensibly similar regions. Pseudo-replication can always be achieved by par-
titioning a single region into two or more sub-regions. Genuine replication
is more common in fields such as neuroanatomy, where the natural spatial
sampling unit is a single field of view under a microscope, and there is a
well-established tradition of using hierarchical sampling designs of the form:
multiple fields of view within tissue sections; multiple tissue sections within
subjects; multiple subjects within experimental treatment groups. Studies of
this kind lend themselves to design-based inference as an alternative to the
more widely prevailing model-based inference for spatial point patterns. See,
for example, Diggle, Lange and Benes (1991), Baddeley et al. (1993) or Eglen,
Diggle and Troy (2005).

1.3 Edge-effects
Edge-effects arise in spatial point pattern analysis when, as is often the case
in practice, the region A on which the pattern is observed is part of a larger
region on which the underlying process operates. The essential difficulty is
then that unobserved events outside A may interact with observed events
within A but, precisely because the events in question are not observed, it is
difficult to take proper account of this.
For some kinds of exploratory analysis, edge-effects can safely be ignored.
We shall discuss when and why this is so at appropriate points in the text.
More generally, we can distinguish between three broad approaches to han-
dling edge-effects: the use of buffer zones; explicit adjustments to take account
of unobserved events; and, when A is rectangular, wrapping A onto a torus
by identifying opposite edges. We will illustrate each of these approaches by
considering a statistic that arises in several contexts, namely the number of
events that occur within a specified distance of an arbitrary event or location.
The buffer zone method consists of carrying out all aspects of the statistical
analysis after conditioning on the locations of all events which fall within a
buffer zone B consisting of all points less than a specified distance, d0 say, from
the edge of A. Let C = A − B denote the remainder of A after subtracting
the buffer zone. Then, it is clear that for any event or location x ∈ C, the
observed number of events within a distance d of x must equal the actual
Introduction 9

number of events in the underlying process within distance d of x, provided

d ≤ d0 , whereas for d > d0 the observed number may be less than the actual
number, thereby biasing any estimates based on these observed numbers. The
choice of d0 in the buffer zone method is awkward, since too small a value
leaves residual edge-effects, whereas too large a value effectively throws away
data unnecessarily. However, the method can be applied in adaptive form,
varying the value of d0 according to the particular statistic being used.
The adjustment method operates by making an “on average” adjustment
for the unobserved events outside A. Again using our simple example to illus-
trate, if we count the observed number, n say, of events within distance d of
a location x, and a(d) denotes the area of intersection between A and a disc
of radius d centred on the location x, then an intuitively sensible estimate
of the actual number of events within distance d of x is nπd2 /a(d). The ad-
justment method is attractive because it makes full use of the observed data,
especially when relatively large-scale effects are of interest. Note, however,
that the adjustments typically lead to an increased sampling variance for the
edge-adjusted estimator by comparison with its unadjusted counterpart. In
essence, this is an example of the common trade-off in statistical estimation
between bias and variance, as edge-corrections seek to eliminate bias at the
expense of some increase in variance.
Toroidal wrapping of a rectangular A is not so much an edge-correction
method as a trick to eliminate edge-effects in particular circumstances. Its
most common use is as a convenient way of simulating realisations of various
kinds of point process. For example, suppose that we wish to simulate a point
process model for the cell centre data shown in Figure 1.3, the most obvious
feature of which is that no two events can occur too close together. If we
attempt to simulate a process of this kind directly on a unit square A, then
points near the edge of A are favoured over points near the centre of A as
potential locations, because of the absence of potentially inhibiting events
outside A. By simulating the process on a torus and subsequently unwrapping
to a unit square for presentation, we avoid this effect. Note that it will seldom
make sense to wrap observed data onto a torus for analysis; for example, if
we were to do this with the cell centre data, then we would observe some very
small toroidal distances between pairs of cells, thus distorting the inhibitory
nature of the underlying process.

1.4 Complete spatial randomness

The hypothesis of complete spatial randomness (henceforth CSR) for a spatial
point pattern asserts that (i) the number of events in any planar region A with
area |A| follows a Poisson distribution with mean λ|A|; (ii) given n events xi
in a region A, the xi are an independent random sample from the uniform dis-
10 Statistical Methods for Spatial and Spatio-Temporal Point Processes

• • •
•• • • ••
• • • •
• •
•
•• • • ••
• •
••
• •• • • •••
• • • • •
• • •• • •• • • ••
• ••
• • •
• • •
•• • • • • •
•• • • • ••• •
•
• ••• •
• •
• • • • • •
• ••
• • • •
•

FIGURE 1.6
Realisation of CSR: 100 events in a unit square.

tribution on A. The self-consistency of (i) and (ii) is not immediately obvious,

but will be established in Chapter 4. In (i), the constant λ is the intensity, or
mean number of events per unit area. According to (i), CSR therefore implies
that the intensity of events does not vary over the plane. According to (ii),
CSR also implies that there are no interactions amongst the events. For ex-
ample, the independence assumption in (ii) would be violated if the existence
of an event at x either encouraged or inhibited the occurrence of other events
in the neighbourhood of x. In developing tests of CSR for sparsely sampled
patterns the starting point will be property (i), whilst for mapped patterns it
is more usual to start with (ii), i.e. to analyse the pattern conditional on the
observed number of events.
Intuitive ideas about what constitutes a “random pattern” can be mislead-
ing. Figure 1.6 shows a realization of 100 events independently and uniformly
distributed on the unit square. Any visual impression of aggregation is illusory.
Note also the superficial similarity to Figure 1.1.
Our interest in CSR is that it represents an idealized standard which, if
strictly unattainable in practice, may nevertheless be tenable as a convenient
Introduction 11

first approximation. Most analyses begin with a test of CSR, and there are
several good reasons for this. Firstly, a pattern for which CSR is not rejected
scarcely merits any further formal statistical analysis. Secondly, tests are used
as a means of exploring a set of data, rather than because rejection of CSR
is of intrinsic interest. Greig-Smith, in the discussion of Bartlett (1971), em-
phasized that ecologists often know CSR to be untenable but nevertheless use
tests of CSR as aids to the formulation of ecologically interesting hypotheses
concerning pattern and its genesis. Thirdly, CSR acts as a dividing hypothesis
to distinguish between patterns which are broadly classifiable as “regular” or
“aggregated”.
Another use of CSR is as a building block in the construction of more
complex models. We shall return to this topic in Chapter 6.

1.5 Objectives of statistical analysis

In any particular application, the objectives of a statistical analysis should
be determined by the scientist’s objectives in collecting the data in question.
We have already given reasons for beginning an analysis with a test of CSR.
What to do next will vary according to context.
In sparse sampling exercises, a specific objective may be to estimate the
intensity. For example, in forestry surveys an important quantity to be esti-
mated is the “stocking density,” or number of stems per hectare. The nature of
the pattern might then be of interest only in so much as it affects the sampling
distribution of the estimator.
With mapped data, the scientist will usually want a more detailed descrip-
tion of the observed pattern than can be provided by a test of CSR. One way
to achieve this is by formulating a parametric stochastic model and fitting it
to the data. If a model can be found that fits the data well, the estimated val-
ues of its parameters provide summary statistics that can be used to compare
ostensibly similar data-sets. More ambitiously, a fitted model can provide an
explanation of the underlying scientific processes. But this must involve an
element of non-statistical inference: quite apart from the obvious fact that
a model which fits the data is not necessarily correct in any absolute sense,
we shall see in Chapter 4 that a simple stochastic model for a spatial point
pattern may admit more than one scientific interpretation.
Of course, it is generally the case that modelling itself is only a means to a
wider end. A well-formulated, and well-fitting, model provides a parsimonious
description of a complex pattern, and one which will be especially useful if
its parameters can be related to scientific hypotheses about the underlying
phenomenon being studied.
Model-fitting is particularly difficult for very heterogeneous data-sets. In
such cases, it may be unhelpful to force a parametric analysis based on unten-
12 Statistical Methods for Spatial and Spatio-Temporal Point Processes

able assumptions. For example, a generic problem in environmental epidemi-

ology is to estimate the spatial variation in the risk of a particular disease,
using data on the locations of individual cases in a geographical region. One
approach to this problem might be to formulate an idealised model for the ob-
served spatial pattern of cases under the assumption that risk is spatially con-
stant, and to investigate deviations of the observed pattern from this model.
An alternative, which would be more in line with classical epidemiological
methods, would be to make a nonparametric comparison between the pattern
of cases and a second pattern of controls, defined to be a random sample from
the population at risk. We shall discuss this idea in detail in Chapter 9.

1.6 The Dirichlet tessellation

Given n distinct events xi in a planar region A, we can assign to xi a “terri-
tory” consisting of that part of A which is closer to xi than to any other xj .
This construction, referred to either as the Dirichlet tessellation or Voronoi
tessellation of the events in A, has been incorporated into stochastic models
of natural phenomena such as inter-plant competition. In these models, plants
whose territories abut are assumed to be in direct competition for available
nutrient; see, for example, Cormack (1979, pp. 171–175). For large n, the tes-
sellation is also the basis of a computationally efficient solution to a number
of problems involving the calculation of distances between events.
Except possibly along the boundary of A, each territory or cell is a convex
polygonal region. Events xi and xj whose cells share a common boundary seg-
ment are said to be contiguous. Typically, each cell vertex is common to three
cells, and the lines joining the pairs of contiguous events define a triangula-
tion of the xi , called the Delaunay triangulation. Thus, cell boundaries can be
obtained as the perpendicular bisectors of the edges of the triangulation, and
cell vertices are the corresponding circumcentres. Figure 1.7 shows a simple
example of both the tessellation and the triangulation associated with n = 12
events in a unit square. Rogers (1964) discusses the mathematical properties
of the Dirichlet tessellation in a general p-dimensional setting.
The construction of the Dirichlet tessellation and the associated Delaunay
triangulation rapidly becomes a non-trivial exercise as n increases. Green and
Sibson (1978) give a remarkably efficient algorithm whose computational cost
increases roughly as n1.5 .
Introduction 13

1.0
•
•
•
0.8
•
•
0.6

•
0.4

•
0.2

•
• • •
0.0

0.0 0.2 0.4 0.6 0.8 1.0

FIGURE 1.7
The Dirichlet tessellation (——) and Delaunay triangulation (– – –) associated
with 12 points in a unit square.

1.7 Monte Carlo tests

Even simple stochastic models for spatial point patterns lead to intractable
distribution theory, and in order to test models against data we shall make
extensive use of Monte Carlo tests (Barnard, 1963).
Quite generally, let u1 be the observed value of a statistic U and let ui : i =
2, ..., s be corresponding values generated by independent random sampling
from the distribution of U under a simple hypothesis H. Let u(j) denote the
j th largest amongst ui : i = 1, ..., s Then, under H,

P {u1 = u(j) } = s−1 , j = 1, ..., s,

and rejection of H on the basis that u1 ranks kth largest or higher gives an
exact, one-sided test of size k/s . This assumes that the values of the ui are all
different, so that the ranking of u1 is unambiguous. If U is a discrete random
variable, for example a count, tied values are possible and we then adopt the
conservative rule of choosing the least extreme rank for u1 . The extension to
two-sided tests is clear.
14 Statistical Methods for Spatial and Spatio-Temporal Point Processes

Hope (1968) gives a number of examples to show that the loss of power
resulting from a Monte Carlo implementation is slight, so that s need not
be very large. For a one-sided test at the conventional 5% level, s = 100 is
adequate.
Power loss is related to Marriott’s (1979) investigation of “blurred critical
regions”, which arise because a value of u1 which would be declared significant
in a classical test may not be declared significant in a Monte Carlo test, and
vice versa. Let the (unknown) distribution function of U under H be F (u).
For a one-sided 5% test with s = 20k, the probability that we reject H, given
that U = u1 , is


k−1
s−1

p(u1 ) = {1 − F (u1 )}r {F (u1 )}s−1−r . (1.1)
r=0
r

For a classical test, represented here by the limit s → ∞, p(u1 ) is 1 or 0

accordingly as F (u1 ) is greater or less than 0.95. The effect of the “blurring”
introduced by (1.1) is measured by Marriott’s Table 1, here reproduced as
Figure 1.8. Marriott concludes also that the extent of the blurring depends
primarily on k, so that if s = 100 is judged to be acceptable for a test at the
5% level, then s = 500 should be used for a test at the 1% level, and pro rata
for tests at smaller levels. These recommended values of s are smaller than
would be considered adequate for the estimation of P {U > u1 | H}. This is
essentially a consequence of the blurring effect. It is the rank of u1 , and not
u1 itself, which is the test statistic. A Monte Carlo test may fail to reject H
when a classical test would have done so, representing a loss of power, but
may also reject H when a classical test would have done so, representing a
gain in power.
A technical point concerning the use of Monte Carlo tests is that in practice
random sampling will be replaced by pseudo-random sampling. In the past,
we have used the generator supplied in the NAG (1977) subroutine library
or a Fortran implementation of the Wichmann and Hill (1982) generator, but
we now routinely use the built-in R function runif() (Venables and Ripley,
1994). Other references for users who wish to understand the theory under-
lying pseudo-random number generators include Kennedy and Gentle (1980)
or Ripley (1987).
A more subtle, but potentially more important, criticism is that Monte
Carlo tests encourage “data-dredging”, since the user can choose the statistic
U to focus on any seemingly aberrant feature of their data. Whilst we admit
that this is a danger, it should be obvious that “significant” results based on
pathological test statistics are of no practical value.
An inherent weakness of the Monte Carlo approach is its restriction to
simple hypotheses H. Composite hypotheses can be tested if pseudo-random
sampling is made conditional on the observed values of sufficient statistics for
any unknown parameters, but this is seldom feasible. Note that a goodness-
of-fit test that ignores the effects of estimating parameters will tend to be
Introduction 15

1.0
0.8
P(reject|u)
0.4 0.6
0.2
0.0

0.70 0.75 0.80 0.85 0.90 0.95 1.00

F (u)

FIGURE 1.8
Blurred critical regions for one-sided, 5% Monte Carlo tests with s = 20, 40,
100 and s → ∞ (adapted from Marriott, 1979).

conservative. This particular difficulty does not arise with tests of CSR for
mapped data, because the observed number of events n is sufficient for the
intensity λ, and conditional on n CSR is a simple hypothesis. But it does
affect the assessment of goodness-of-fit for more general stochastic models.
An approximate remedy, which we discuss further in Chapter 6, is to measure
goodness-of-fit by a statistic that is not directly related to the procedure used
to estimate the parameters of the model.
The principal advantage to be set against the above is that the investigator
need not be constrained by known distribution theory, but rather can and
should use informative statistics of their own choosing.
When asymptotic distribution theory is available, Monte Carlo testing
provides an exact alternative for small samples and a useful check on the ap-
plicability of the asymptotic theory. If the results of classical and Monte Carlo
tests are in substantial agreement, little or nothing has been lost; if not, the
16 Statistical Methods for Spatial and Spatio-Temporal Point Processes

explanation is usually that the classical test uses inappropriate distributional

assumptions.

1.8 Software
Spatial point pattern analysis is computationally intensive, not least because
of the heavy reliance on Monte Carlo methods of inference. As noted in the
Preface, R has become the computing environment of choice for many statisti-
cians. The splancs package (Rowlingson and Diggle, 1993) gives a wide range
of functions for statistical analysis of spatial point patterns. The Spatstat li-
brary, written by Adrian Baddeley and Rolf Turner, also implements a wide
range of methods, with a stronger emphasis than splancs on parametric mod-
elling. Many of the analyses reported in this book were implemented using a
combination of splancs, Spatstat and Voronoi (a package for computation
of the Dirichlet tessellation, written by Rolf Turner), together with some ad-
ditional functions written by the author.
More sophisticated displays than those shown in this book, for example
colour-coded overlays of point pattern maps and contour maps, can most
easily be produced using a Geographical Information System (GIS). A wide
variety of commercial and open-source GIS packages are now available. Also,
a number of R packages have been written to provide GIS-like functionality
within the R environment. For a detailed description of spatial data-handling
in R, see for example Bivand, Pebesma and Gomez-Rubio (2008).
References

Alexander, F.E. and Boyle, P. (1996). Methods for Investigating Localized

Clustering of Disease. Lyon : International Agency for Research on Cancer.
Anderson, N.H. and Titterington, D.M. (1997). Some methods for investigat-
ing spatial clustering, with epidemiological applications. Journal of the Royal
Statistical Society, A 160, 87–105.
Ardern, K. (2001). Report on the Possible Increase in Cancer Cases in North
Liverpool and Potential Links to the Site of the Former Incinerator at Fazak-
erley Hospital. Liverpool : Liverpool Health Authority.
Baddeley, A.J. (1999). Spatial sampling and censoring. In Stochastic Geome-
try: Likelihood and Computation, eds. O.E. Barndorff-Nielsen, W.S. Kendall
and M.N.M. Van Lieshout, 37–78. London : Chapman and Hall.
Baddeley, A.J. and van Lieshout, M.N.M. (1995). Area-interaction point pro-
cesses. Annals of the Institute of Statistical Mathematics, 47, 601–19.
Baddeley, A.J. and Møller, J. (1989). Nearest-neighbour Markov point pro-
cesses and random sets. International Statistical Review, 57, 89–121.
Baddeley, A.J., Møller, J. and Waagepetersen, R. (2000). Non- and semi-
parametric estimation of interaction in inhomogeneous point patterns. Statis-
tica Neerlandica, 54, 329–50.
Baddeley, A.J., Moyeed, R.A., Howard, C.V. and Boyde, A. (1993). Analysis
of a three-dimensional point pattern with replication. Applied Statistics, 42,
641–68.
Baddeley, A.J. and Silverman, B.S. (1984). A cautionary example on the use of
second-order methods for analyzing point patterns. Biometrics, 40, 1089–93.
Baddeley, A. and Turner, R. (2000). Practical maximum pseudolikelihood for
spatial point patterns (with Discussion). Australian and New Zealand Journal
of Statistics, 42, 283–322.
Bagchi, R., Henrys, P.A., Brown, P.E., Burslem, D.F.R.P., Diggle, P.J., Gu-
natilleke, C.V.C., Gunatilleke, I.A.U., Kassim, A.R., Law, R. and Valencia,
R.L. (2011). Spatial patterns reveal negative density dependence and habitat
associations in tropical trees. Ecology, 92, 1723-1729

245
246 References

Bailey, T.C. and Gatrell, A.G. (1995). Interactive Spatial Data Analysis. Har-
low : Longman.
Barnard, G. A. (1963). Contribution to the discussion of Professor Bartlett’s
paper. Journal of the Royal Statistical Society. B 25, 294.
Bartlett, M. S. (1937). Properties of sufficiency and statistical tests. Proceed-
ings of the Royal Society, A 160, 268–282.
Bartlett, M. S. (1964). Spectral analysis of two-dimensional point processes.
Biometrika, 51, 299–311.
Bartlett, M. S. (1971). Two-dimensional nearest neighbour systems and their
ecological applications. In Statistical Ecology, Vol. 1, eds. G.P. Patil, E.C.
Pielou and W.E. Waters, 179–84. University Park: Pennsylvania State Uni-
versity Press.
Bartlett, M. S. (1975). The Statistical Analysis of Spatial Pattern. London:
Chapman and Hall.
Bell, M.L. and Grunwald, G.K. (2004). Mixed models for the analysis of repli-
cated spatial point patterns. Biostatistics, 5, 633–648.
Berman, M. and Diggle, P. (1989). Estimating weighted integrals of the
second-order intensity of a spatial point process. Journal of the Royal Sta-
tistical Society, B 51, 81–92.
Berman, M. and Turner, T.R. (1992). Approximating point process likelihoods
with GLIM. Applied Statistics, 41, 31–8.
Bernal, J. D. (1960). Geometry of the structure of monatomic liquids. Nature.
185, 68–70.
Besag, J. (1974). Spatial interaction and the statistical analysis of lattice sys-
tems (with discussion). Journal of the Royal Statistical Society, B 34, 192–236.
Besag, J. (1975). Statistical analysis of non-lattice data. The Statistician, 24,
179–95.
Besag, J. (1977). Contribution to the discussion of Dr Ripley’s paper. Journal
of the Royal Statistical Society, B 39, 193–5.
Besag, J. (1978). Some methods of statistical analysis for spatial data. Bulletin
of the International Statistical Institute, 47, 77–92.
Besag, J. and Diggle, P. J. (1977). Simple Monte Carlo tests for spatial pattern.
Applied Statistics, 26, 327–33.
Besag, J. and Gleaves, J. T. (1973). On the detection of spatial pattern in
plant communities. Bulletin of the International Statistical Institure, 45 (1),
153–8.
References 247

Besag, J., Milne, R. and Zachary, S. (1982). Point process limits of lattice
processes. Journal of Applied Probability, 19, 210–6.
Besag, J. and Newell, J. (1991). The detection of clusters of rare diseases.
Journal of the Royal Statistical Society, A 154, 143–55.
Bivand, R.S., Pebesma, E.J. and Gomez-Rubio, V. (2008). Applied Spatial
Data Analysis with R. New York: Springer
Bostoen, K., Chalabi, Z. and Grais, R.F. (2007). Optimisation of the T-square
sampling method to estimate population sizes. Emerging Themes in Epidemi-
ology, 4, doi:10.1186/1742-7622-4-7.
Breslow, N.E. and Day, N.E. (1980). Statistical Methods in Cancer Research,
Volume 1: The Analysis of Case-Control Studies. Lyon : International Agency
for Research on Cancer.
Brown, D. (1975). A test of randomness of nest spacing. Wildfowl 26, 102-3.
Brown, D. and Rothery, P. (1978). Randomness and local regularity of points
in a plane. Biometrika, 65, 115–22.
Brown, P.E., Karesen, K., Roberts, G.O. and Tonellato, S. (2000). Blur-
generated non-separable space-time models. Journal of the Royal Statistical
Society, B 62, 847–60.
Brown, S. and Holgate, P. (1974). The thinned plantation. Biometrika, 61,
253–62.
Brown, T. (1979). Position dependent and stochastic thinning of point pro-
cesses. Stochastic Processes and their Applications, 9, 189–93.
Brown, T. C., Silverman, B. W. and Milne, R.K. (1981). A class of two-type
point processes. Z. Wahrscheinlichkeitstheorie verw. Geb. 58, 299–308.
Byth, K. (1982). On robust distance-based intensity estimators. Biometrics
38, 127–35.
Byth, K, and Ripley, B. D. (1980). On sampling spatial patterns by distance
methods. Biometrics 36, 279–284.
Catana, A. J. (1963). The wandering quarter method of estimating population
density. Ecology 44, 349–60.
Chetwynd, A.G. and Diggle, P.J. (1998). On estimating the reduced second
moment measure of a stationary spatial point process. Australian and New
Zealand Journal of Statistics, 40, 11–5.
Chetwynd, A.G., Diggle, P.J., Marshall, A. and Parslow, R. (2001). Inves-
tigation of spatial clustering from individually matched case-control studies.
Biostatistics, 2, 277–93
Chilès, J-P and Delfiner, P. (1999). Geostatistics. Hoboken: Wiley.
248 References

Chilès, J-P and Delfiner, P. (2012). Geostatistics, 2nd edn. Hoboken: Wiley.
Clark, P. J. and Evans, F. C. (1954). Distance to nearest neighbour as a
measure of spatial relationships in populations. Ecology, 35, 23–30.
Cliff, A. D. and Ord, J. K. (1981). Spatial Processes: Models and Applications.
London: Pion.
Cook-Mozaffari, P., Darby, S., Doll, R., Forman, D., Hermon, C. and Pike,
M.C. (1989). Geographical variation in mortality from leukaemia and other
cancers in England and Wales in relation to proximity to nuclear installations,
1969–78. British Journal of Cancer, 59, 476–85.
Cormack, R.M. (1977). The invariance of Cox and Lewis’ statistic for the
analysis of spatial patterns. Biometrika 64, 143–4.
Cormack, R.M. (1979). Spatial aspects of competition between individuals. In
Spatial and Temporal Analysis in Ecology, eds. R.M. Cormack and J. K. Ord,
152–211. Fairland: International Co-operative Publishing House.
Cottam, G. and Curtis, J. T. (1949). A method for making rapid surveys of
woodlands, by means of pairs of randomly selected trees. Ecology 30, 101–4.
Cox, D.R. (1955). Some statistical methods related with series of events (with
discussion). Journal of the Royal Statistical Society. B 17, 129–64.
Cox, D.R. (1972a). The statistical analysis of dependencies in point processes.
In Stochastic Point Processes, ed. P. A. W. Lewis, 55–66. New York: Wiley.
Cox, D.R. (1972b). Regression models and life tables (with Discussion). Jour-
nal of the Royal Statistical Society B, 34, 187–220.
Cox, D.R. (1977). The role of significance tests. Scandinavian Journal of
Statistics, 4, 49–70.
Cox, D.R. and Isham, V. (1980). Point Processes. London: Chapman and Hall.
Cox, D.R. and Lewis, P. A. W. (1966). The Statistical Analysis of Series of
Events. London: Methuen.
Cox, D.R. and Lewis, P. A. W. (1972). Multivariate point processes. Proceed-
ings of the 6th Berkeley Symposium in Mathematical Statistics and Probability,
3, 401-448.
Cox, T. F. (1976). The robust estimation of the density of a forest stand using
a new conditioned distance method. Biometrika 63, 493–500.
Cox, T. F. (1979). A method for mapping the dense and sparse-regions of a
forest stand. Applied Statistics, 28, 14–9.
Cox, T. F. and Lewis, T. (1976). A conditioned distance ratio method for
analysing spatial patterns. Biometrika 63, 483–92.
Cressie, N.A.C. (1991). Statistics for Spatial Data. New York: Wiley.
References 249

Cressie, N. and Huang, H-C. (1999). Classes of non-separable spatio-temporal

stationary covariance functions. Journal of the American Statistical Associa-
tion, 94, 1330–1340.
Cressie, N. and Wikle, C.K. (2011). Statistics for Spatio-Temporal Data. Hobo-
ken: Wiley.
Crick, F. H. C. and Lawrence, P. A. (1975). Compartments and polychones
in insect development. Science 189, 340–7.
Cuzick, J. and Edwards, R. (1990). Spatial clustering for inhomogeneous pop-
ulations (with Discussion). Journal of the Royal Statistical Society, B 52,
73–104.
Daley, D. J. and Vere-Jones, D. (1972). A summary of the theory of point
processes. In Stochastic Point Processes, ed. P.A.W. Lewis, 299–383. New
York: Wiley.
Daley, D. J. and Vere-Jones, D. (2002). Introduction to the Theory of Point
Processes: Elementary Theory and Methods. New York : Springer.
Daley, D. J. and Vere-Jones, D. (2005). Introduction to the Theory of Point
Processes: General Theory and Structure. New York : Springer.
Diggle, P. J. (1975). Robust density estimation using distance methods.
Biometrika, 62, 39–48.
Diggle, P. J. (1977a). A note on robust density estimation for spatial point
patterns. Biometrika, 64, 91–5.
Diggle, P. J. (1977b). The detection of random heterogeneity in plant popu-
lations. Biometrics, 33, 390–4.
Diggle, P. J. (1978). On parameter estimation for spatial point processes.
Journal of the Royal Statistical Society, B 40, 178–81.
Diggle, P. J. (1979a). On parameter estimation and goodness-of-fit testing for
spatial point patterns. Biometrics, 35, 87–101.
Diggle, P. J. (1979b). Statistical methods for spatial point patterns in ecology.
In Spatial and Temporal Analysis in Ecology, eds. R.M. Cormack and J. K.
Ord, 95–150. Fairland: International Co-operative Publishing House.
Diggle, P. J. (1981a). Some graphical methods in the analysis of spatial point
patterns. In Interpreting Multivariate Data, ed. V. Barnett, 55–73. Chichester:
Wiley.
Diggle, P. J. (1981b). Statistical analysis of spatial point patterns. New
Zealand Statistician, 16, 22–41.
Diggle, P.J. (1985a). Displaced amacrine cells in the retina of a rabbit: analysis
of a bivariate spatial point pattern. Journal of Neuroscience Methods, 18, 115–
25.
250 References

Diggle, P.J. (1985b). A kernel method for smoothing point process data. Ap-
plied Statistics, 34, 138–47.
Diggle, P.J. (1990). A point process modelling approach to raised incidence
of a rare phenomenon in the vicinity of a prespecified point. Journal of the
Royal Statistical Society, A 153, 349–62.
Diggle, P.J. (2006). Spatio-temporal point processes, partial likelihood, foot-
and-mouth. Statistical Methods in Medical Research, 15, 325–336.
Diggle, P. L, Besag, J. and Gleaves, J. T. (1976). Statistical analysis of spatial
point patterns by means of distance methods. Biometrics, 32, 659–67.
Diggle, P.J. and Cox, T.F. (1983). Some distance-based tests of independence
for sparsely-sampled multivariate spatial point patterns. International Statis-
tical Review, 51, 11–23.
Diggle, P.J., Eglen, S.J. and Troy, J.B. (2006). Modelling the bivariate spatial
distribution of amacrine cells. In Case Studies in Spatial Point Processes, ed
A Baddeley, P. Gregori, J. Mateu, R. Stoica and D. Stoyan, 215–233. New
York: Springer.
Diggle, P. Elliott, P., Morris, S. and Shaddick, G. (1997). Regression modelling
of disease risk in relation to point sources. Journal of the Royal Statistical
Society, A 160, 491–505.
Diggle, P.J., Fiksel, T., Grabarnik, P., Ogata, Y., Stoyan, D. and Tanemura,
M. (1994). On parameter estimation for pairwise interaction point processes.
International Statistical Review, 62, 99–117.
Diggle, P.J., Gates, D.J. and Stibbard. A. (1987). A non-parametric estimator
for pairwise-interaction point processes. Biometrika, 74, 763–70.
Diggle, P.J, Gomez-Rubio, V., Brown,P.E., Chetwynd, A.G and Gooding, S.
(2007). Second-order analysis of inhomogeneous spatial point process data.
Biometrics, 63, 550–557.
Diggle, P.J. and Gratton, R.J. (1984) Monte Carlo methods of inference for
implicit statistical models (with discussion). Journal of the Royal Statistical
Society B, 46, 193–227.
Diggle, P.J., Kaimi, I. and Abellana, R. (2010). Partial likelihood analysis of
spatio-temporal point process data. Biometrics 66, 347–354.
Diggle, P.J., Knorr-Held, L., Rowlingson, B., Su, T., Hawtin, P. and Bryant, T.
(2003). Towards on-line spatial surveillance. In Monitoring the Health of Pop-
ulations: Statistical Methods for Public Health Surveillance., eds. R. Brook-
meyer and D. Stroup. Oxford : Oxford University Press.
Diggle, P.J., Lange, N. and Benes, F.M. (1991). Analysis of variance for repli-
cated spatial point patterns in clinical neuroanatomy. Journal of the American
Statistical Association, 86, 618–25.
References 251

Diggle, P. J. and Matérn, B. (1981). On sampling designs for the estima-

tion of point-event nearest neighbour distributions. Scandinavian Journal of
Statistics, 7, 80–4.
Diggle, P.J., Mateu, J. and Clough, H.E. (2000). A comparison between para-
metric and non-parametric approaches to the analysis of replicated spatial
point patterns. Advances in Applied Probability, 32, 331–43
Diggle, P.J., Menezes, R. and Su, T-L. (2010). Geostatistical analysis under
preferential sampling (with discussion). Applied Statistics 59, 191–232.
Diggle, P. J. and Milne, R.K. (1983a). Negative binomial quadrat counts and
point processes. Scandinavian Journal of Statistics, 10, 257–67.
Diggle, P. J. and Milne, R.K. (1983b). Bivariate Cox processes: some models
for bivariate spatial point patterns. Journal of the Royal Statistical Society. B
45, 11–21.
Diggle, P.J., Moraga, P., Rowlingson, B. and Taylor, B. (2013). Spatial
and spatio-temporal log-Gaussian Cox processes: extending the geostatisti-
cal paradigm. Statistical Science (to appear)
Diggle, P.J., Morris, S.E. and Wakefield, J.C. (2000). Point-source modelling
using matched case-control data. Biostatistics, 1, 89–105.
Diggle, P.J. and Ribeiro, P.J. (2007). Model-based Geostatistics. New York:
Springer.
Diggle, P.J. and Rowlingson, B.S. (1994). A conditional approach to point
process modelling of elevated risk. Journal of the Royal Statistical Society, A
157, 433-40.
Diggle, P.J., Zheng, P. and Durr, P. (2005). Non-parametric estimation of
spatial segregation in a multivariate point process. Applied Statistics, 54, 645–
58.
Donnelly, K. (1978). Simulations to determine the variance and edge-effect of
total nearest neighbour distance. In Simulation Methods in Archaeology, ed.
I. Hodder, 91–5. London: Cambridge University Press.
Douglas, J. B. (1979). Analysis with Standard Contagious Distributions. Fair-
land: International Co-operative Publishing House.
Duczmal, L., Kulldorff, M. and Huang, L. (2006). Evaluation of spatial scan
statistics for irregularly shaped clusters. Journal of Computational and Graph-
ical Statistics, 15, 428–442.
Du Rietz, G. E. (1929). The fundamental units of vegetation. Proceedings of
the International Congress of Plant Science, Ithaca, 1, 623–7.
Eberhardt, L. L. (1967). Some developments in “distance sampling”. Biomet-
rics, 23, 207–16.
252 References

Efron, B. and Tibshirani, R.J. (1993). An Introduction to the Bootstrap. Lon-

don : Chapman and Hall.
Eglen, S.J., Diggle, P.J. and Troy, J.B. (2005). Homotypic constraints domi-
nate positioning of on- and off-center beta retinal ganglion cells. Visual Neu-
roscience, 22, 859–871.
Elliott, P., Beresford, J.A., Jolley, D.J., Pattenden, S.H. and Hills, M. (1992).
Cancer of the larynx and lung near incinerators of waste solvents and oils in
Britain. In Geographical and Environmental Epidemiology: methods for small-
area studies, eds. P. Elliott, J. Cuzick, D. English and R. Stern, 359–67. Oxford
: Oxford University Press.
Elliott, P., Wakefield, J.C., Best, N.G. and Briggs, D.J. (2000). Spatial Epi-
demiology: methods and applications. Oxford : Oxford University Press.
Evans, D. A. (1953). Experimental evidence concerning contagious distribu-
tions in ecology. Biometrika, 40, 186–211.
Fanshawe, T.F. and Diggle, P.J. (2011). Bivariate geostatistical modelling: a
review and an application to spatial variation in radon concentrations. Envi-
ronmental and Ecological Statistics 19, 139–160.
Feller, W. (1968). An Introduction to Probability Theory and Its Applications,
Vol. 1, 3rd edn. New York: Wiley.
Finkenstadt, B., Held, L. and Isham, V. (2007). Statistical Methods for Spatio-
Temporal Systems. Boca Raton: Chapman and Hall/CRC.
Fisher, R.A. (1925). Statistical Methods for Research Workers. Edinburgh :
Oliver and Boyd.
Fisher, R.A. (1935). The Design of Experiments. Edinburgh : Oliver and Boyd.
Fisher, R.A., Thornton, H. G. and Mackenzie, W. A. (1922). The accuracy
of the plating method of estimating the density of bacterial populations, with
particular reference to the use of Thornton’s agar medium with soil samples.
Annals Applied Botany, 9, 325–59.
Gabriel, E. and Diggle, P.J. (2009). Second-order analysis of inhomogeneous
spatio-temporal point process data. Statistica Neerlandica 63, 43–51
Gardner, M.J. (1989). Review of reported incidences of childhood cancer rates
in the vicinity of nuclear installations in the U.K. Journal of the Royal Sta-
tistical Society, A 152, 307–25.
Gates, D.J. and Westcott, M. (1986). Clustering estimates in spatial point
distributions with unstable potentials. Annals of the Institute of Statistical
Mathematics, 38 A, 55-67.
Gelfand, A., Diggle, P.J., Fuentes, M. And Guttorp, P. (2010). Handbook of
Spatial Statistics. Boca Raton: CRC Press.
References 253

Gerrard, D. J. (1969). Competition quotient: a new measure of the competi-

tion affecting individual forest trees. Research Bulletin No. 20, Agricultural
Experiment Station, Michigan State University.
Geyer C.J. (1999). Likelihood inference for spatial point processes. In Stochas-
tic Geometry, Likelihood and Computation, eds. O.E. Barndorff-Nielsen, W.S.
Kendall and M.N.M. van Lieshout, 79-140. London: Chapman and Hall.
Geyer, C.J. and Thompson, E.A. (1992). Constrained Monte Carlo maximum
likelihood for dependent data (with discussion). Journal of the Royal Statis-
tical Society, B 54, 657–99.
Ghent, A. W. (1963). Studies of regeneration of forest stands devastated by
spruce budworm. Forest Science, 9, 295–310.
Gilks, W.R, Richardson, S. and Spiegelhalter, D.J. (1996). Markov Chain
Monte Carlo in Practice. London : Chapman and Hall.
Gill, P. E. and Murray, W. (1972). Quasi-Newton methods for unconstrained
optimization. Journal of the Institute of Mathematics and its Applications, 9,
91–108.
Gneiting, T. (2002). Nonseparable, stationary covariance functions for space-
time data. Journal of the American Statistical Association, 97, 590–600.
Gneiting, T. and Guttorp, P. (2010). Continuous parameter spatio-temporal
processes. In Handbook of Spatial Statistics, eds. A.E. Gelfand, P.J. Diggle, M.
Fuentes, P. Guttorp, 427–436. Boca Raton: Chapman and Hall/CRC Press.
Goodall, D. G. (1965). Plot-less tests of inter-specific association. Journal of
Ecology, 53, 197–210.
Green, P. J. and Sibson, R. (1978). Computing Dirichlet tessellations in the
plane. Computer Journal, 21, 168–73.
Greenland, S. and Morgenstern, H. (1989). Ecological bias, confounding and
effect modification. International Journal of Epidemiology, 18, 269–74.
Greig-Smith, P. (1952). The use of random and contiguous quadrats in the
study of the structure of plant communities. Annals of Botany, 16, 293–316.
Greig-Smith, P. (1964). Quantitative Plant Ecology, 2nd edn. London: Butter-
worth.
Greig-Smith, P. (1979). Pattern in vegetation. Journal of Ecology, 67, 755–79.
Groendyke, C., Welch, D. and Hunter, D.R. (2012). A network-based analysis
of the 1861 Hagelloch measles data. Biometrics, 68(3), 755-765.
Hanisch, K. H. and Stoyan, D. (1979). Formulas for the second-order analysis
of marked point processes. Math. Operationsforsch. Statist. Ser. Statist., 10,
555–60.
254 References

Harkness, R.D. and Isham, V. (1983). A bivariate spatial point pattern of

ants’ nests. Applied Statistics, 32, 293–303.
Hastie, T.J. and Tibshirani, R.J (1990). Generalized Additive Models. London
: Chapman and Hall.
Heikkinen, J. and Penttinen, A. (1999). Bayesian smoothing in the estimation
of the pair potential function of Gibbs point processes. Bernoulli, 5, 1119–36.
Hernández, A. and Ruiz, X. (2003). Predation on common tern eggs by the
yellow-legged gull at the Ebro Delta. Scientia Marina, 67, 95-101.
Hines, W.G.S. and Hines, R.J.0. (1979). The Eberhardt index and the detec-
tion of non-randomness of spatial point distributions. Biometrika, 66, 73–80.
Ho, L. P. and Stoyan, D. (2008). Modelling marked point patterns by intensity-
marked Cox processes. Statistics and Probability Letters, 78, 11941199.
Hodder, I. and Orton, C. (1976). Spatial Analysis in Archaeology. London:
Cambridge University Press.
Hogmander, H. and Sarrka, A. (1999). Multitype spatial point patterns with
hierarchical interactions. Biometrics, 55, 1051–8.
Holgate, P. (1964). The efficiency of nearest neighbour estimators. Biometrics,
20, 647–9.
Holgate, P. (1965a). Tests of randomness based on distance methods.
Biometrika, 52, 345–53.
Holgate, P. (1965b). Some new tests of randomness. Journal of Ecology, 53,
261–6.
Holgate, P. (1965c). The distance from a random point to the nearest point
of a closely packed lattice. Biometrika, 52, 261–3.
Holgate, P. (1972). The use of distance methods for the analysis of spatial
distributions of points. In Stochastic Point Processes, ed. P.A.W. Lewis, 122–
53. New York: Wiley.
Hope, A. C. A. (1968). A simplified Monte Carlo significance test procedure.
Journal of the Royal Statistical Society, B 30, 582–98.
Hopkins, B. (1954). A new method of determining the type of distribution of
plant individuals. Annals of Botany, 18, 213–26.
Hsuan, F. (1979). Generating uniform polygonal random pairs. Applied Statis-
tics, 28, 170–2.
Hughes, A. (1981). Cat retina and the sampling theorem: the relation of tran-
sient and sustained brisk-unit cut-off frequency to α and β-mode cell density.
Experimental Brain Research, 42, 196–202.
Hutchings, M. J. (1979). Standing crop and pattern in pure stands of Mercu-
References 255

rialis perennis and Rubus fruticosus in mixed deciduous woodland. Oikos, 31,
351–7.
Ilian, J., Penttinen, A., Stoyan, H. and Stoyan, D. (2008). Statistial Analysis
and Modelling of Spatial Point Patterns. Chichester: Wiley.
Jalava, K., Jones, J.A., Goodchild, T., Clifton-Hadley, R., Mitchell, A., Story,
A. and Watson, J.M. (2007). No increase in human cases of Mycobacterium
bovis disease despite resurgence of infections in cattle in the United Kingdom
Epidemiology and Infection, 135, 40–45.
Jarner, M.F., Diggle, P.J. and Chetwynd, A.G. (2002). Estimation of spatial
variation in risk using matched case-control data. Biometrical Journal (sub-
mitted).
Kathirgamatamby, N. (1953). Note on the Poisson index of dispersion.
Biometrika, 40,225–8.
Keeling, M.J., Woolhouse, M.E.J., Shaw, D.J., Matthews, L., Chase-Topping,
M., Haydon, D.T., Cornell, S.J., Kappey, J., Wilesmith, J. and Grenfell, B.T.
(2001). Dynamics of the 2001 UK foot and mouth epidemic: stochastic dis-
persal in a heterogeneous landscape. Science, 294, 813–7.
Kelly, F. P. and Ripley, B. D. (1976). A note on Strauss’ model for clustering.
Biometrika, 63, 357–60.
Kelsall, J.E. and Diggle, P.J. (1995a). Kernel estimation of relative risk.
Bernoulli, 1, 3–16.
Kelsall, J.E. and Diggle, P.J. (1995b). Nonparametric estimation of spatial
variation in relative risk. Statistics in Medicine, 14, 2335–42.
Kelsall, J.E. and Diggle, P.J. (1998). Spatial variation in risk: a nonparametric
binary regression approach. Applied Statistics 47, 559–73.
Kendall, M. G. (1970). Rank Correlation Methods, 4th edn. London: Griffin.
Kennedy, W. J. and Gentle, J. E. (1980). Statistical Computing. New York:
Marcel Dekker.
Kershaw, K. A. (1957). The use of cover and frequency in the detection of
pattern in plant communities. Ecology, 38, 291–9.
Kershaw, K. A. (1973). Quantitative and Dynamic Plant Ecology, 2nd edn.
London: Arnold.
Kingman, J. F. C. (1977). Remarks on the spatial distribution of a reproducing
population. Journal of Applied Proabability, 14, 577–83.
Knox, G. (1963). Detection of low intensity epidemicity. British Journal of
Preventive and Social Medicine, 17, 121–7.
256 References

Knox, G. (1964). Epidemiology of childhood leukaemia in Northumberland

and Durham. British Journal of Preventive and Social Medicine, 18, 17–24.
Kulldorff, M. (1997). A spatial scan statistic. Communications in Statistics
Theory and Methods, 26, 1481–96.
Kulldorff, M. (1999). An isotonic spatial scan statistic for geographical disease
surveillance. Journal of the National Institute of Public Health, 48, 94–101.
Kulldorff1, M., Huang, L., Pickle, L. and Duczmal, L. (2006). An elliptic
spatial scan statistic Statistics in Medicine 25, 3929-3943.
Kulldorff, M., Farzad Mostashari,F., Duczmal, L., Yih, W.K.1, Kleinman,
K. and Platt, R. (2007). Multivariate scan statistics for disease surveillance.
Statistics in Medicine 26, 1824-1833.
Lawson, A.B (1989). Discussion on cancer near nuclear installations. Journal
of the Royal Statistical Society, A 152, 374–5.
Lawson, A.B. (1993). On the analysis of mortality events associated with a
prespecified fixed point. Journal of the Royal Statistical Society, A 156, 363–
77.
Lawson, A. and Leimich, P. (2000). Approaches to the space-time mod-
elling of infectious disease behaviour. IMA Journal of Mathematics Applied in
Medicine and Biology 17, 1–13.
Lewis, P. A. W. and Shedler, G. S. (1979). Simulation of non-homogenous
Poisson processes by thinning. Naval Research Logistics Quarterly, 26, 403-
13.
Lloyd, M. (1967). Mean crowding. Journal of Animal Ecology, 36, 1–30.
Lotwick H. W. (1981). PhD thesis, University of Bath.
Lotwick, H. W. and Silverman, B. W. (1982). Methods for analysing spatial
processes of several types of points. Journal of the Royal Statistical Society,
B 44, 406–13.
McCullagh, P. and Nelder, J.A. (1989). Generalized Linear Models 2nd edn.
London : Chapman and Hall.
Ma, C. (2003). Families of spatio-temporal covariance models. Journal of Sta-
tistical Planning and Inference, 116, 489–501.
Ma, C. (2008). Recent developments in the construction of spatio-temporal
covariance models. Stochastic Environmental Research and Risk Assessment,
22, S39–S47.
Mannion, D. (1964). Random space-filling in one dimension. Publications of
the Mathematical Institute of the Hungarian Academy of Science, 9, 143–53.
References 257

Marriott, F. H. C. (1979). Monte Carlo tests: how many simulations? Applied

Statistics, 28, 75–7.
Matérn, B. (1960). Spatial Variation. Meddelanden fran statens skogs-
forsningsinstitut, Vol. 49 (5). Stockholm: Statens Skogsforsningsinstitut.
Matérn, B. (1971). Doubly stochastic Poisson processes in the plane. In Statis-
tical Ecology, Vol. 1, eds. G.P. Patil, E.C. Pielou and W.E. Waters, 195–213.
University Park: Pennsylvania State University Press.
Matérn, B. (1986) Spatial Variation. New York: Springer.
Matheron, G. (1975). Random Sets and Integral Geometry. New York: Wiley.
Maynard-Smith, J. (1974). Models in Ecology. London: Cambridge University
Press.
Mead, R.(1974). A test for spatial pattern at several scales using data from a
grid of contiguous quadrats. Biometrics, 30, 295–307.
Møller, J. and Sorensen, M. (1994). Statistical analysis of a spatial birth-and-
death process model with a view to modelling linear dune fields. Scandinavian
Journal of Statistics, 21, 1–19.
Møller, J., Syversveen, A.R. and Waagepetersen, R.P. (1998). Log-Gaussian
Cox processes. Scandinavian Journal of Statistics, 25, 451–82.
Møller, J. and Waagepetersen, R.P. (2004). Statistical Inference and Simula-
tion for Spatial Point Processes. London : Chapman and Hall.
Moore, P. G. (1954). Spacing in plant populations. Ecology 35, 222–7.
Morisita, M. (1959). Measuring the dispersion of individuals and analysis of
the distributional patterns. Memoirs of the Faculty of Science, Kyushu Uni-
versity Series E (Biology), 2, 215–35.
NAG (1977). Fortran Library Manual. Oxford: NAG Executive.
Nelder, J. A. and Mead, R. (1965). A simplex method for function minimiza-
tion. Computer Journal, 7, 308–13.
Neyman, J. (1939). On a new class of contagious distributions, applicable in
entomology and bacteriology. Annals of Mathematical Statistics, 10, 35–57.
Neyman, J. and Scott, E. L. (1958). Statistical approach to problems of cos-
mology (with discussion). Journal of the Royal Statistical Society, B 20, 1–43.
Numata, M. (1961). Forest vegetation in the vicinity of Choshi. Coastal flora
and vegetation at Choshi, Chiba Prefecture IV. Bulletin of Choshi Marine
Laboratory, Chiba University, 3, 28–48 [in Japanese].
Ogata, Y. and Tanemura, M. (1981). Estimation of interaction potentials of
spatial point patterns through the maximum likelihood procedure. Annals of
the Institute of Statistical Mathematics, B 33, 315–38.
258 References

Ogata, Y. and Tanemura, M. (1984). Likelihood analysis of spatial point pat-

terns. Journal of the Royal Statistical Society, B 46, 496–518.
Ogata, Y. and Tanemura, M. (1986). Likelihood estimation of interaction po-
tentials and external fields of inhomogeneous spatial point patterns. In Pacific
Statistical Congress, eds. I.S.Francis, B.J.F.Manly and F.C.Lam, 150–4. Am-
sterdam, Elsevier.
Ohser, J. and Stoyan, D. (1981). On the second-order and orientation analysis
of planar stationary point processes. Biometrical Journal, 23, 523–33.
Ord, J. K. (1978). How many trees in a forest? Mathematical Scientist, 3,
23–33.
Ord, J. K. (1979). Time-series and spatial patterns in ecology. In Spatial and
Temporal Analysis in Ecology, eds. R.M. Cormack and J.K. Ord, 1–94. Fair-
land: International Co-operative Publishing House.
Patil, S. A., Burnham, K. P. and Kovner, J. L. (1979). Non-parametric esti-
mation of plant density by the distance method. Biometrics 35, 613–22.
Peebles, P. J. E. (1974). The nature of the distribution of galaxies.Astronomy
and Astrophysics, 32,197–202.
Penttinen, A. (1984). Modelling Interaction in Spatial Point Patterns: pa-
rameter estimation by the maximum likelihood method. Jyvaskyla Studies in
Computer Science, Economics and Statistics, 7. University of Jyvaskyla.
Perry, J. N. and Mead, R. (1979). On the power of the index of dispersion test
to detect spatial pattern. Biometrics 35, 613–22.
Persson, 0. (1964). Distance methods. Studia Forestalia Suecica 15, 1–68.
Persson, 0. (1971). The robustness of estimating density by distance measure-
ments. In Statistical Ecology, Vol. 2, eds. G.P. Patil, E.C. Pielou and W.E.
Waters, 175–90. University Park: Pennsylvania State University Press.
Pielou, E. C. (1977). Mathematical Ecology. New York: Wiley.
Pollard, J. H. (1971). On distance estimators of density in randomly dis-
tributed forests. Biometrics, 27, 991–1002.
Preston, C. J. (1977). Spatial birth-and-death processes. Bulletin of the In-
ternational Statistical Institute, 46 (2), 371–91.
Prince, M., Chetwynd, A., Diggle, P., Jarner, M., Metcalf, J. and James, O.
(2001). The geographical distribution of primary biliary cirrhosis in a well-
defined cohort. Hepatology 34, 1083–8.
Rathbun, S.L. (1996). Estimation of Poisson intensity using partially observed
concomitant variables. Biometrics, 52, 226–42.
References 259

Ripley, B. D. (1976). The second-order analysis of stationary point processes.

Journal of Applied Proabability, 13, 255–66.
Ripley, B. D. (1977). Modelling spatial patterns (with discussion). Journal of
the Royal Statistical Society, B 39, 172–212.
Ripley, B. D. (1978). Spectral analysis and the analysis of pattern in plant
communities. Journal of Ecology, 66, 965–81.
Ripley, B. D. (1979a). Tests of “randomness” for spatial point patterns. Jour-
nal of the Royal Statistical Society, B 41, 368–74.
Ripley, B. D. (1979b). Simulating spatial patterns: dependent samples from a
multivariate density. Applied Statistics, 28, 109–12.
Ripley, B.D. (1981). Spatial Statistics. New York: Wiley.
Ripley, B.D. (1987). Stochastic Simulation. New York: Wiley.
Ripley, B.D. (1988). Statistical Inference for Spatial Processes. Cambridge :
Cambridge University Press.
Ripley, B. D. and Kelly, F. P. (1977). Markov point processes. Journal of the
London Mathematical Society, 15, 188–92.
Ripley, B. D. and Silverman, B. W. (1978). Quick tests for spatial regularity.
Biometrika, 65, 641–2.
Rogers, C. A. (1964). Packing and Covering. London: Cambridge University
Press.
Rosenblatt, M. (1956). Remarks on some nonparametric estimates of a density
function. Annals of Mathematical Statistics, 27, 832–7.
Rodrigues, A. and Diggle, P.J. (2010). A class of convolution-based models
for spatio-temporal processes with non-separable covariance structure. Scan-
dinavian Journal of Statistics, 37, 553–567.
Rodrigues, A., Diggle, P. and Assuncao, R. (2010). Semiparametric approach
to point source modelling in epidemiology and criminology. Applied Statistics,
59, 533–542.
Rowlingson, B.S. and Diggle, P.J. (1993). Splancs: Spatial point pattern anal-
ysis code in S-plus. Computers in Geosciences, 19, 627–55.
Rue, H. and Held, L. (2005). Gaussian Markov Random Fields: Theory and
Applications. London: CRC Press.
Rue, H., Martino, S. and Chopin, N. (2009). Approximate Bayesian inference
for latent Gaussian models using integrated nested Laplace approximations
(with discussion). Journal of the Royal Statistical Society B, 71, 319-392.
Sarkka, A. (1983). Pseudo-likelihood approach for pair potential estimation
260 References

of Gibbs processes. Jyvaskyla Studies in Computer Science, Economics and

Statistics, 22. University of Jyvaskyla.
Schabenberger and Gotway, C.A. (2004). Statistical Methods for Spatial Data
Analysis. Boca Raton: Chapman and Hall/CRC.
Selvin, H.C. (1958). Durkheim’s ‘suicide’ and problems of empirical research.
American Journal of Sociology, 63, 607–19.
Schlather, M.S., Ribeiro, P.J. and Diggle, P.J. (2004). Detecting dependence
between marks and locations of marked point processes. Journal of the Royal
Statistical Society, B 66, 79–93.
Silverman, B. W. (1981). Density estimation for univariate and bivariate data.
In Interpreting Multivariate Data, ed. V. Barnett, 37–53. Chichester: Wiley.
Silverman, B.W. (1986). Density estimation for statistics and data analysis.
London: Chapman and Hall.
Silverman, B. W. and Brown, T. (1978). Short distances, flat triangles and
Poisson limits. Journal of Applied Proabability, 15, 815–25.
Skellam, J. G. (1958). On the derivation and applicability of Neyman’s Type
A distribution. Biometrika, 45, 32–6.
Sprent, P. (1981). Quick Statistics. Harmondsworth: Penguin.
Stein, M.L. (1991). A new class of estimators for the reduced second moment
measure of point processes. Biometrika, 78, 281–6.
Stiteler, W. M. and Patil, G. P. (1971). Variance to mean ratio and Morisita’s
index as measures of spatial pattern in ecological populations. In Statistical
Ecology, Vol. 1, eds. G.P. Patil, E.C. Pielou and W.E. Waters, 423–59. Uni-
versity Park: Pennsylvania State University Press.
Stoyan, D. (1979). Interrupted point processes. Biometrical Journal, 21, 607–
10.
Stoyan, D., Kendall, W.S. and Mecke, J. (1995). Stochastic Geometry and its
Applications, 2nd edn. New York : Wiley.
Stoyan, D. and Stoyan, H. (1994). Fractals, Random Shapes and Point Fields.
New York : Wiley.
Strauss, D. J. (1975). A model for clustering. Biometrika, 62, 467–75.
Tanemura, M. (1979). On random complete packing by discs. Annals of the
Institute of Statistical Mathematics, 31, 351–65.
Tango, T. (1995). A class of tests for detecting ‘general’ and ‘focused’ cluster-
ing of rare diseases. Statistics in Medicine, 14, 2323–34.
Tango, T. (2000). A test for spatial disease clustering adjusted for multiple
testing. Statistics in Medicine, 19, 191–204.
References 261

Taylor, B.M., Davies, T.M., Rowlingson, B.S. and Diggle, P.J. (2013). lgcp:
Inference with spatial and spatio-temporal log-Gaussian Cox processes in R.
Journal of Statistical Software 52(4) www.jstatsoft.org/v52/i04
Thomas, M. (1949). A generalization of Poisson’s binomial limit for use in
ecology. Biometrika 36, 18–25.
Thompson, H.R. (1956). Distribution of distance to nth nearest neighbour in
a population of randomly distributed individuals. Ecology, 37, 391–4.
Tildesley, M.J., Bessell, P.R., Keeling, M.J. and Woolhouse, M.E.J. (2009).
The role of pre-emptive culling in the control of foot-and-mouth disease. Pro-
ceedings of the Royal Society B 276, 3239–3248.
Townsend, P., Phillimore, P. and Beattie, Q.A. (1988). Health and Depriva-
tion: Inequalities and the North. London : Croom Helm.
Upton, G.J.G. and Fingleton, B. (1985). Spatial Data Analysis by Example,
Volume 1: Point Pattern and Quantitative Data. Chichester: Wiley.
Upton, G.J.G. and Fingleton, B. (1989). Spatial Data Analysis by Example,
Volume 2: Categorical and Directional Data. Chichester: Wiley.
Van Lieshout, M.N.M. (2000). Markov Point Processes and their Applications.
London : Imperial College Press.
Van Lieshout, M.N.M. and Baddeley, A.J. (1996). A nonparametric measure
of spatial interaction in point processes. Statistical Neerlandica, 50, 344–61.
Venables, W.N. and Ripley, B.D. (1994). Modern Applied Statistics with S-
Plus. New York: Springer.
Waller, L. and Gotway, C.A. (2004). Applied Spatial Statistics for Public
Health Data. New York: Wiley.
Warren, W. G. (1971). The centre satellite concept as a basis for ecological
sampling. In Statistical Ecology, Vol. 2, eds. G.P. Patil, E.C. Pielou and W.E.
Waters, 87–116. University Park: Pennsylvania State University Press.
Warren, W. G. and Batcheler, C. L. (1979). The density of spatial patterns:
robust estimation through distance methods. In Spatial and Temporal Analysis
in Ecology, eds. R.M. Cormack and J. K. Ord, 240–70. Fairland: International
Co-operative Publishing House.
Wichmann, B.A. and Hill, I.D. (1982). Algorithm AS183. An efficient and
portable pseudo-random number generator. Applied Statistics, 31, 188–90.
(Correction, 33, 123)
Williams, J.R., Alexander, F.E., Cartwright, R.A. and McNally, R.J.Q. (2001).
Methods for eliciting aetiological clues from geographically clustered cases of
disease, with application to leukaemia-lymphoma data. Journal of the Royal
Statistical Society, A 164, 49–60.
262 References

Wood, S.N. (2006). Generalized Additive Models: An Introduction with R.

Boca Raton: Chapman and Hall.
Woodward, M. (1999). Epidemiology: Study Design and Data Analysis. Lon-
don : Chapman and Hall.
Zhuang, J., Ogata, Y. and Vere-Jones, D. (2002). Stochastic Declustering of
space-time earthquake occurrences. Journal of the American Statistical Asso-
ciation, 97, 369–380.
This page intentionally left blank