Download the full version of the textbook now at textbookfull.

com

Asymptotic Analysis of Mixed Effects Models:

Theory, Applications, and Open Problems 1st
Edition Jiming Jiang

https://textbookfull.com/product/asymptotic-
analysis-of-mixed-effects-models-theory-
applications-and-open-problems-1st-edition-jiming-
jiang/

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

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

Linear and Generalized Linear Mixed Models and Their

Applications 2nd Edition Jiming Jiang

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

textbookfull.com

Biota Grow 2C gather 2C cook Loucas

https://textbookfull.com/product/biota-grow-2c-gather-2c-cook-loucas/

textbookfull.com

Longitudinal Data Analysis Autoregressive Linear Mixed

Effects Models Ikuko Funatogawa

https://textbookfull.com/product/longitudinal-data-analysis-
autoregressive-linear-mixed-effects-models-ikuko-funatogawa/

textbookfull.com

The Art of Cinematic Storytelling: A Visual Guide to

Planning Shots, Cuts, and Transitions Kelly Gordon Brine

https://textbookfull.com/product/the-art-of-cinematic-storytelling-a-
visual-guide-to-planning-shots-cuts-and-transitions-kelly-gordon-
brine/
textbookfull.com
Ganong’s Physiology Examination and Board Review 1st
Edition Kim E. Barrett

https://textbookfull.com/product/ganongs-physiology-examination-and-
board-review-1st-edition-kim-e-barrett/

textbookfull.com

Finance Book The Understand the numbers even if you re not

a finance professional Financial Times 1st Edition Warner

https://textbookfull.com/product/finance-book-the-understand-the-
numbers-even-if-you-re-not-a-finance-professional-financial-times-1st-
edition-warner/
textbookfull.com

Data Structures & Algorithms in Python John Canning

https://textbookfull.com/product/data-structures-algorithms-in-python-
john-canning/

textbookfull.com

The Pie Book Over 400 Classic Recipes Louis P. De Gouy

https://textbookfull.com/product/the-pie-book-over-400-classic-
recipes-louis-p-de-gouy/

textbookfull.com

Gray Hat Hacking The Ethical Hacker s Handbook 5th Edition

Shon Harris Daniel Regalado Allen Harper Harris Shon
Regalado Daniel Harper Allen
https://textbookfull.com/product/gray-hat-hacking-the-ethical-hacker-
s-handbook-5th-edition-shon-harris-daniel-regalado-allen-harper-
harris-shon-regalado-daniel-harper-allen/
textbookfull.com
Greek tragedy on the move : the birth of a Panhellenic art
form c. 500-300 BC First Edition Edmund Stewart

https://textbookfull.com/product/greek-tragedy-on-the-move-the-birth-
of-a-panhellenic-art-form-c-500-300-bc-first-edition-edmund-stewart/

textbookfull.com
Monographs on Statistics and Applied Probability 155

Asymptotic Analysis of
Mixed Effects Models
Theory, Applications,
and Open Problems

Jiming Jiang
University of California, Davis, USA
 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)
129. Constrained Principal Component Analysis and Related Techniques Yoshio Takane (2014)
130. Randomised Response-Adaptive Designs in Clinical Trials Anthony C. Atkinson and Atanu Biswas (2014)
131. Theory of Factorial Design: Single- and Multi-Stratum Experiments Ching-Shui Cheng (2014)
132. Quasi-Least Squares Regression Justine Shults and Joseph M. Hilbe (2014)
133. Data Analysis and Approximate Models: Model Choice, Location-Scale, Analysis of Variance, Nonparametric
Regression and Image Analysis Laurie Davies (2014)
134. Dependence Modeling with Copulas Harry Joe (2014)
135. Hierarchical Modeling and Analysis for Spatial Data, Second Edition Sudipto Banerjee, Bradley P. Carlin,
and Alan E. Gelfand (2014)
136. Sequential Analysis: Hypothesis Testing and Changepoint Detection Alexander Tartakovsky, Igor Nikiforov,
and Michèle Basseville (2015)
137. Robust Cluster Analysis and Variable Selection Gunter Ritter (2015)
138. Design and Analysis of Cross-Over Trials, Third Edition Byron Jones and Michael G. Kenward (2015)
139. Introduction to High-Dimensional Statistics Christophe Giraud (2015)
140. Pareto Distributions: Second Edition Barry C. Arnold (2015)
141. Bayesian Inference for Partially Identified Models: Exploring the Limits of Limited Data Paul Gustafson (2015)
142. Models for Dependent Time Series Granville Tunnicliffe Wilson, Marco Reale, John Haywood (2015)
143. Statistical Learning with Sparsity: The Lasso and Generalizations Trevor Hastie, Robert Tibshirani, and
Martin Wainwright (2015)
144. Measuring Statistical Evidence Using Relative Belief Michael Evans (2015)
145. Stochastic Analysis for Gaussian Random Processes and Fields: With Applications Vidyadhar S. Mandrekar and
Leszek Gawarecki (2015)
146. Semialgebraic Statistics and Latent Tree Models Piotr Zwiernik (2015)
147. Inferential Models: Reasoning with Uncertainty Ryan Martin and Chuanhai Liu (2016)
148. Perfect Simulation Mark L. Huber (2016)
149. State-Space Methods for Time Series Analysis: Theory, Applications and Software
Jose Casals, Alfredo Garcia-Hiernaux, Miguel Jerez, Sonia Sotoca, and A. Alexandre Trindade (2016)
150. Hidden Markov Models for Time Series: An Introduction Using R, Second Edition
Walter Zucchini, Iain L. MacDonald, and Roland Langrock (2016)
151. Joint Modeling of Longitudinal and Time-to-Event Data
Robert M. Elashoff, Gang Li, and Ning Li (2016)
152. Multi-State Survival Models for Interval-Censored Data
Ardo van den Hout (2016)
153. Generalized Linear Models with Random Effects: Unified Analysis via H-likelihood, Second Edition
Youngjo Lee, John A. Nelder, and Yudi Pawitan (2017)
154. Absolute Risk: Methods and Applications in Clinical Management and Public Health
Ruth M. Pfeiffer and Mitchell H. Gail (2017)
155. Asymptotic Analysis of Mixed Effects Models: Theory, Applications, and Open Problems
Jiming Jiang (2017)
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2017 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

Version Date: 20170510

International Standard Book Number-13: 978-1-4987-0044-3 (Hardback)

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

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

For permission to photocopy or use material electronically from this work, please access
www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc.
(CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization
that provides licenses and registration for a variety of users. For organizations that have been granted
a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To Thuan
 Contents

List of Figures 13

List of Tables 15

Preface 17

About the Author 19

1 Introduction 1
1.1 Example: Generalized estimating equations 1
1.2 Example: REML estimation 4
1.3 Example: Salamander mating data 6
1.4 A roadmap 8

2 Asymptotic Analysis of Linear Mixed Models 11

2.1 Maximum likelihood 12
2.2 Restricted maximum likelihood 17
2.2.1 Asymptotic behavior of REML: Earlier studies 17
2.2.2 Asymptotic behavior of REML: Asymptotic superiority 19
2.2.3 Asymptotic behavior of REML: Wald consistency 25
2.3 Asymptotic covariance matrix in non-Gaussian LMM 29
2.3.1 Empirical method of moments 30
2.3.2 Partially observed information 33
2.4 Mixed model prediction 38
2.4.1 Best linear unbiased prediction 40
2.4.2 Observed best prediction 43
2.4.3 Classified mixed model prediction 47
2.5 Linear mixed model diagnostics 52
2.6 Misspecified mixed model analysis 58
2.6.1 A misspecified LMM and MMMA 59
2.6.2 Preliminaries 61
2.6.3 Consistency and asymptotic distribution 64
2.6.4 Further empirical results 66
2.7 Additional bibliographical notes and open problems 69
2.7.1 Bibliographical notes 69
2.7.2 Open problems 71

9
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
10 CONTENTS
3 Asymptotic Analysis of Generalized Linear Mixed Models 75
3.1 Asymptotic deficiency of PQL 77
3.2 Generalized estimating equations 79
3.2.1 Estimating equations and GEE 79
3.2.2 Iterative estimating equations 84
3.2.3 Method of simulated moments 90
3.2.4 Robust estimation in GLMM 97
3.3 Maximum likelihood estimation 102
3.3.1 Historical notes 103
3.3.2 The subset argument 105
3.3.3 Two types of consistency; extensions 107
3.3.4 Example 109
3.4 Conditional inference 111
3.4.1 Maximum posterior 111
3.4.2 Conditional inference 115
3.4.3 Quadratic inference function 119
3.5 Maximization by parts 122
3.6 GLMM diagnostics 124
3.6.1 Tailoring 126
3.6.2 Application to GLMM 128
3.6.3 A simulated example 131
3.7 Additional bibliographical notes and open problems 133
3.7.1 Bibliographical notes 133
3.7.2 Open problems 135

4 Small Area Estimation 139

4.1 The Prasad–Rao method 139
4.2 The Fay–Herriot model 143
4.3 Extension of Prasad–Rao method 150
4.4 Empirical best prediction with binary data 158
4.5 Resampling methods in SAE 166
4.5.1 Jackknifing the MSPE of EBP 166
4.5.2 Monte-Carlo jackknife for SAE after model selection 175
4.5.3 Bootstrapping mixed models 186
4.6 Additional bibliographical notes and open problems 197
4.6.1 Bibliographical notes 197
4.6.2 Open problems 199

5 Asymptotic Analysis in Other Mixed Effects Models 203

5.1 Nonlinear mixed effects models 203
5.2 Preliminaries 212
5.3 Frailty models 217
5.3.1 Asymptotic analysis of gamma-frailty models 218
5.3.2 Asymptotic analysis of proportional hazards mixed-effects
models 221
CONTENTS 11
5.4 Joint modeling of survival and longitudinal data 223
5.5 Additional bibliographical notes and open problems 228
5.5.1 Bibliographical notes 228
5.5.2 Open problems 230

References 233

Index 249
 List of Figures

2.1 Heritability–REML provides the right answer despite model mis-

specification. 62
2.2 Heritability–REML provides the right answer despite model mis-
specification. 67
2.3 Heritability–REML provides the right answer despite model mis-
specification. 68

3.1 Theoretical versus Empirical Distributions: n = 100. Left: pdfs;

Right: cdfs. 132
3.2 Theoretical versus Empirical Distributions: n = 200. Left: pdfs;
Right: cdfs. 133

4.1 Boxplots of coverage probabilities. Nominal Level: 95%. From left

to right: Coverage probabilities of prediction intervals constructed
by 1–true MSPE; 2–P-R MSPE estimator; and 3–converting the
second-order unbiased Log-MSPE estimator. 178
4.2 Boxplots of %RB when Model 1 is the true model. In each plot, from
left to right: 1–naive estimator, 2–bootstrap estimator, and 3–McJack
estimator, of log-MSPE. 184
4.3 Boxplots of %RB when Model 2 is the true model. In each plot, from
left to right: 1–naive estimator, 2–bootstrap estimator, and 3–McJack
estimator, of log-MSPE. 185
4.4 Boxplots of %RB. In each plot, from left to right: 1–DHM, 2–
bootstrap, 3–McJack. Scales are different due to the huge difference
in range. 187

13
 List of Tables

2.1 RDT versus Jackknife—Size 38

2.2 RDT versus Jackknife—Power (Nominal Level 0.05) 38
2.3 Average MSPE for Prediction of Mixed Effects. 52

3.1 Simulated Mean and Standard Error 91

3.2 Comparison of Estimators 96
3.3 Simulation Results: Mixed Logistic Model 99
3.4 Simulation Results: Beta-Binomial 100
3.5 Empirical and Theoretical Quartiles 132
3.6 Simulated Size and Power 133

4.1 Summary Statistics for Coverage Probabilities (Nominal Level 95%) 177
4.2 Log-MSPE Estimation 186
4.3 DHM versus McJack in %RB 188
4.4 Coverage Probabilities and Lengths of Prediction Intervals 191
4.5 Simplified Adaptive Fence: Simulation Results 194

15
 Preface

Large sample techniques play fundamental roles in all fields of statistics. As quoted
from [1], “Large-sample techniques provide solutions to many practical problems;
they simplify our solutions to difficult, sometimes intractable problems; they justify
our solutions; and they guide us to directions of improvements.” Mixed effects mod-
els, including linear mixed models, generalized linear mixed models (GLMMs), non-
linear mixed effects models, and non-parametric mixed effects models, are complex
models by nature, yet, these models are extensively used in practice. See, for exam-
ple, [2], [3], [4]. For such reasons, there have been substantial growing interests in
finding, simplifying, and justifying the solutions for this important class of models.
This monograph provides a comprehensive account on asymptotic analysis of
mixed effects models. Historically, large sample theory has provided guidelines for
methodology developments as well as applications of these models. For example, re-
stricted maximum likelihood (REML) has established itself as the preferred method
of linear mixed model (LMM) analysis largely due to an asymptotic theory that has
been developed in the 1990s. As another example, penalized quasi-likelihood (PQL)
has been gradually giving way to (full) maximum likelihood (ML) inference par-
tially due to inconsistency of the PQL estimators that has been discovered; on the
other hand, major computational developments, such as Monte-Carlo E-M algorithm
and data cloning, have made the ML inference feasible.
A large portion of the monograph is related to the author’s research publications
over the past 20 years on linear and generalized linear mixed models. These include
the author’s work in the 1990s on REML asymptotics; the author’s 15-year pursuit in
proving consistency of the ML estimator (MLE) in GLMM involving crossed random
effects; the author’s long-standing interest in small area estimation (SAE; e.g., [5]);
and other topics of mixed model analysis such as mixed model diagnostics and mixed
model selection. Of course, the monograph also covers research work of many other
authors in this area. In particular, there has been growing interest in semi-parametric
and non-parametric mixed effects models and related asymptotics. Asymptotic the-
ory has been used as guidelines, for example, in the analysis of longitudinal data,
where semi/non-parametric mixed effects models have been used. Another area in
which extensive asymptotic studies have been done is estimation of mean squared
prediction errors (MSPE) and construction of prediction intervals in SAE.
The author intends to build a strong connection between the asymptotic theory
and real-life applications. Interesting real-data examples are provided to motivate the
theoretical development. For example, the salamander-mating data [6] has been a
driving force for the computational developments for GLMM, and a primary moti-

17
18 PREFACE
vation for proving the consistency of the MLE for GLMMs involving crossed ran-
dom effects. Some recent advances in genome-wide association study have motivated
asymptotic analysis of misspecified LMM. The estimation of MSPE offers another
excellent example where asymptotic study is highly motivated by practical problems.
In a way, mixed effects model asymptotics are difficult problems due to the com-
plexity of these models and the fact that observations are typically correlated under
these models. As a result, classical limit theory regarding independent random vari-
ables (e.g., [1], Chapter 6) does not apply, or at least cannot be directly used. There
have been some vast developments in the asymptotic theory for mixed effects models
over the past decades. On the other hand, some important, and challenging, problems
regarding the mixed effects model asymptotics remain unsolved. The author provides
a number of such challenges as open problems. A main goal of this monograph is to
encourage young researchers to devote their talents, and energy, to asymptotic stud-
ies that are motivated by practical problems. For this purpose, we have supplemented
eah chapter (starting Chapter 2) with a section of additional bibliographical notes and
open problems. The problems range from theoretical and methodological to compu-
tational and practical issues, and challenges that are important, as well as interesting.
The main topics of this monograph are asymptotics, which are often treated as
theoretical properties. Thus, in a way, we are theoreticians. However, we are theo-
reticians driven by practical problems, and this is made clear throughout the entire
monograph. It should also be noted that the monograph is not written in a theorem-
proof style. Instead, we focus on interpretation of the results as well as assumptions,
and applications to problems of practical interest. Of course, interested readers can
always explore the detailed theorems and proofs in the references provided. In this
regard, we have adopted a strategy similar to [1].
More importantly, asymptotic analyses often guide us to directions of improve-
ments, as noted earlier. The monograph intends to provide a comprehensive account
of such improvements in mixed model analysis.
Finally, it should be noted that the monograph is not a “collection of results.”
Instead of trying to cover all of the relevant results, we focus on important methods
that have had impacts in the fields, have been used in other work, and are potentially
useful in future studies. In spite of our efforts, there is a nonzero probability that
some of the important methods in these categories are missing. We apologize for
such omissions, and promise to catch up in the next edition of the monograph.
The author wishes to thank Professors Ronghui (Lily) Xu, Jimin Ding, and Jian-
hua Huang for communications and suggestions. The author also would like to ex-
press his appreciation to Professors Thuan Nguyen, Can Yang, and Dr. Cecilia Dao
for graphical assistance. In addition, the author is grateful to a number of anonymous
reviewers at various stages of the manuscript for their comments and suggestions.
Jiming Jiang
Davis, California
 About the Author

Jiming Jiang is a professor of statistics at the University of California, Davis. He re-

ceived his Ph. D. in statistics in 1995 from the University of California, Berkeley. His
research interests include asymptotic theory, mixed effects models, model selection,
small area estimation, analysis of longitudinal data, and big data intelligence. He is
the author/co-author of two previously published books and one monograph: Lin-
ear and Generalized Linear Mixed Models and Their Applications (Springer 2007),
Large Sample Techniques for Statistics (Springer 2010), and (with T. Nguyen) The
Fence Methods (World Scientific, 2015). He is an editorial board member of sev-
eral major statistical journals, including The Annals of Statistics and Journal of the
American Statistical Association. He is a Fellow of the American Statistical Asso-
ciation, a Fellow of the Institute of Mathematical Statistics, and an Elected Member
of the International Statistical Institutes. Jiming Jiang is a co-recipient of Outstand-
ing Statistical Application Award (American Statistical Association, 1998), the first
co-recipient of the NISS Former Postdoc Achievement Award (National Institute of
Statistical Sciences, 2015), and a Yangtze River Scholar (Chaired Professor, 2016),
which is widely considered the highest academic honor awarded (to a foreigner) by
the Ministry of Education of the People’s Republic of China.

19
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
 Chapter 1

Introduction

Statistics is nowadays often referred to as data science. Today’s data are much bigger,
and more complex, than what a statistician had to handle decades ago; hence, not
surprisingly, today’s statistical models are also much more complex. From a classical
i.i.d. (independent and identically distributed) model to a linear regression model;
from a linear regression model to a linear mixed model; and from a linear mixed
model to a nonlinear mixed effects model, we are witnessing an ever-changing story
about our data, in good ways.
The complexity of statistical modeling allows practitioners to better understand
the nature of their data, such as correlations among the observations, as more and
more people have come to realize that the independent-data assumption is often un-
realistic. On the other hand, the advances in statistical modeling also present new
challenges. For example, given the importance of large sample techniques in statis-
tics (e.g., [1]), how much do we (really) know about the large sample properties of
an estimator derived under a complex statistical model?
In this monograph, we focus on one particular class of complex statistical mod-
els, known as mixed effects models. A common feature of these models is that the
observations are correlated. The correlation may be modeled by the so-called ran-
dom effects, in the form of a hierarchical model, or in terms of a variance-covariance
structure. The detailed definitions of different types of mixed effects models will be
given in the later chapters, but here we would like to first offer a few examples as a
preview for what we are going to do.

1.1 Example: Generalized estimating equations

In medical studies, repeated measures are often collected from patients over time.
Let yit be the measure collected from patient i at time t. Here 1 ≤ i ≤ m, m being the
total number of patients that are involved, and t ∈ Ti , where Ti is a subset of times
which may be patient-dependent. Let yi = (yit )t∈Ti denote the vector of measures
collected from patient i. Quite often, a vector of explanatory variables, or covariates,
are available for each patient and at each observational time. Let xitk denote the kth
component of the covariates for patient i at time t, 1 ≤ k ≤ p, and Xi = (xitk )t∈Ti ,1≤k≤p
the matrix of covariates for patient i. Data with such kind of layout is often referred
to as longitudinal data, and are encountered in many fields well beyond medical

1
2 INTRODUCTION
studies (e.g., [7]). In general, we shall use the term subject, which in the case of
medical study would correspond to a patient.
A class of models, known as marginal models, for the longitudinal data are de-
fined by assuming that yi , 1 ≤ i ≤ m are independent such that E(yi ) = µi = µi (β ),
where β is a vector of parameters, and Var(yi ) = Vi , where Vi is an unknown co-
variance matrix. Usually, νi is a known function of β that is associated with Xi , for
example, µi = Xi β in case of a linear model. On the other hand, so far Vi is as-
sumed completely unknown except, perhaps, that it is finite, so the assumption on
the variance-covariance structure among the data within the subject is very weak.
However, in many cases, the main interest is the mean function, µi = (µit )t∈Ti . For
example, in a medical study, µit corresponds to the mean turnover of a marker of
bone formation, osteocalcin, for subject i at time t [8]. Nevertheless, the covariance
matrix is closely related to any reasonable measure of uncertainty of an estimator of
the mean function. Thus, at some point, one will have to deal with Vi , even though it
is not of direct interest.
Because µi (·) is a known function, to assess the mean function, one needs to
estimate the parameter vector, β . A well-known method of estimation is called gen-
eralized estimating equations, of GEE, defined by solving the equation
m
∂ µ′
∑ ∂ βi Vi−1(yi − µi) = 0, (1.1)
i=1

where the (k,t) element of ∂ µi′ /∂ β is ∂ µit /∂ βk . The motivation of (1.1) will be
given in Section 3.2. In order to solve (1.1) for β , one needs to know the Vi ’s, which
are often difficult to estimate. One reason is that there are many unknown parameters
involved in Vi , 1 ≤ i ≤ m. In fact, if the Vi ’s are completely unspecified, there are
∑mi=1 ni (ni + 1)/2 unknown parameters involved in all of the Vi ’s, where ni = |Ti |, the
cardinality of Ti . Of course, one may consider modeling the Vi ’s by assuming some
parametric forms for the Vi ’s, and thus reduce the number of unknown parameters
involved. A challenging problem, in the latter regard, is to determine which paramet-
ric form is reasonable to use. Again, this topic will be discussed later. Here, let us
first introduce an idea, first introduced by Liang and Zeger ([9]; also see [10]) that, in
a way, may have surprised many. Let Ṽi be a “working covariance matrix” (WCM),
for each i. Here by WCM it means that Ṽi is not necessarily the true Vi , or even an
estimator of the latter. But, the WCM is known and ready to use for the computation.
For example, the simplest WCM is the identity matrix, Ṽi = Ini , where In denotes
the n-dimensional identity matrix. Liang and Zeger suggested to estimate β by solv-
ing (1.1) with Vi replaced by Ṽi , 1 ≤ i ≤ m. A reader’s first response to the proposal
might be something like “Wait a minute. The Vi ’s in (1.1) are supposed to be the
true covariance matrices; replacing them by the Ṽi ’s, which may not be even close to
the Vi ’s?” Nevertheless, Liang and Zeger [9] provided a solid argument in support of
their proposal by proving that the GEE estimator, obtained by solving (1.1) with Vi
replaced by Ṽi (1 ≤ i ≤ m), is consistent (e.g., [1], p. 33) for any Ṽi satisfying some
mild conditions.
Although the result of Liang and Zeger may sound surprising, it does not mean
that knowing the Vi ’s (or not) does not matter. In fact, it matters, but only to the
EXAMPLE: GENERALIZED ESTIMATING EQUATIONS 3
“second-order”. Roughly speaking, a first-order asymptotic property states that the
estimator is consistent, while a second-order (asymptotic) property states that the
estimator is efficient in the sense that its asymptotic variance is the smallest (e.g.,
[11], Section 6.2). A related issue is that, because the GEE estimator with WCM is
likely to be inefficient, its variance may be large. Therefore, it is important to have a
good assessment of the variance of the GEE estimator. Liang and Zeger [9] proposed
the following asymptotic method, widely known as the “sandwich” estimator. Denote
the right side of (1.1), with Vi replaced by Ṽi , by g(β ), a vector-valued function with
the same dimension as β . Then, the GEE estimator, β̂ , satisfies

∂g
0 = g(β̂ ) ≈ g(β ) + (β̂ − β ),
∂β′

where β denotes the true parameter vector, and the partial derivatives are evaluated
at the true β . Here the approximation is due to the Taylor expansion, and consistency
of β̂ , as noted above. More precisely, the approximation is in the sense that the dif-
ference is of lower order in terms of convergence in probability. See, for example,
Section 3.4 of [1]. One can make a further approximation (in the same sense) by
replacing ∂ g/∂ β ′ by its expectation, and thus obtain the following approximation:
0 ≈ g(β ) + E(∂ g/∂ β ′)(β̂ − β ), or
  −1
∂g
β̂ − β ≈ − E g(β ), (1.2)
∂β′

assuming non-singularity of E(∂ g/∂ β ′ ). By taking the covariance matrix on both

sides of (1.2), one expects that, naturally,
  −1   ′ −1
∂g ∂g
Var(β̂ ) ≈ E Var{g(β )} E . (1.3)
∂β′ ∂β

The right side of (1.3) already looks like a “sandwich,” but it is not yet an estimator.
To convert it to an estimator, write gi = (∂ µi′ /∂ β )Ṽi−1 (yi − µi ), and note that
  !
m
∂g ∂ gi
E = E ∑ , (1.4)
∂β′ i=1 ∂ β
′

m
Var{g(β )} = ∑ Var(gi )
i=1
m
= ∑ E(g2i )
i=1
!
m
= E ∑ g2i . (1.5)
i=1

Here, in deriving (1.5), we have used the fact that the gi ’s are independent with mean
zero, provided that the WCM are constant matrices. We then approximate the left
4 INTRODUCTION
sides of (1.4) and (1.5) by their right sides, respectively, with the expectation signs
removed. The latter is a standard technique used in asymptotics (e.g., [1], Section
3.1). Thus, combined with (1.3), we obtain the following estimator of Var(β̂ ), the
sandwich estimator:
!−1 ! !−1
m
∂ ĝ m m
∂ ĝ ′
c β̂ ) = i 2 i
Var( ∑ ′ ∑ ĝi ∑ ∂ β , (1.6)
i=1 ∂ β i=1 i=1

where ĝi and ∂ ĝi /∂ β ′ are gi and ∂ gi /∂ β ′ , respectively, with β replaced by β̂ . The
sandwich estimator can be used to obtain, for example, the standard errors of β̂k , 1 ≤
k ≤ p by taking the square roots of the diagonal elemenets of Var( c β̂ ).

1.2 Example: REML estimation

REML—restricted or residual maximum likelihood, is a method of estimating disper-
sion parameters, or variance components, involved in a linear model. Here the linear
model typically has a more complicated variance-covariance structure than the linear
regression model, although the latter may serve as one of the simplest examples to
illustrate REML (see below).
Let us begin with the simplest of all. Suppose that Y1 , . . . ,Yn are i.i.d. observations
with mean µ and variance σ 2 . A standard estimator of the variance, σ 2 , is the sample
variance, defined as

1 n
σ̂ 2 = ∑ (Yi − Ȳ )2 , (1.7)
n − 1 i=1

where Ȳ = n−1 ∑ni=1 Yi is the sample mean. If, furthermore, the Yi ’s are normal, an
alternative estimator of σ 2 is the maximum likelihood estimator (MLE), which is

1 n
σ̂ 2 = ∑ (Yi − Ȳ )2 . (1.8)
n i=1

Comparing (1.7) with (1.8), the only difference is the denominator: n − 1 for the
sample variance and n for the MLE. In fact, this is the difference between REML
and maximum likelihood (ML) in this very simple case. In other words, the sample
variance is, indeed, the REML estimator of σ 2 in this case. Of course, if n is large,
there is not much difference relatively between n and n − 1, so the REML estimator
and MLE should be close in values. However, this may not be the case when the
situation is more complicated.
A case that is more general than the i.i.d. situation is when the observations are
independent but not identically distributed. One of such cases is linear regression.
Before we move on, we would like to make a note on the notation we use. Classi-
cal statistics and probability theory (e.g., [11]) were built on the i.i.d. case, where a
standard notation for an observation is the same as for a random variable, which is in
a capital letter, say, Yi . Sometimes, to distinguish a random variable from a realized
Random documents with unrelated
content Scribd suggests to you:
sinking steamer was an episode, not a disaster. In few words,
Captain O’Shea assured them that he had no intention of letting this
uncomfortable little happening interfere with the business for which
he had employed them. The insurance underwriters would be out of
pocket, but who cared a rap for them, anyhow? Thereupon he
issued orders, swiftly, intelligently, with masterful vehemence. The
two boats which appeared most serviceable were swung outboard
and held ready to launch. They would hold a dozen men each
without crowding. Water-kegs were filled, the galley and store-room
ransacked for tins of meat and biscuit, bags of potatoes and rice.
The fire-arms and cutlasses were served out and the cases of
ammunition divided between the two boats. Meanwhile the Whang
Ho continued to sink with a certain dignity and decorum. One could
find nothing dramatic in this shipwreck. Every one moved with haste,
but there was no outcry.
Only one mischance marred the exodus from the Whang Ho. All
hands were absorbed, and quite naturally, in delaying their
departure as little as possible. Delay meant something worse than
wet feet. In fact, the main deck was almost level with the water
when the boats were ready to shove clear. For once the Whang Ho
had moved rapidly, although in a lamentable direction. With so much
to do in so short a time, it was not extraordinary that the vigilant
espionage which surrounded Charley Tong Sin should be relaxed, not
to say forgotten, for a moment. Even Captain O’Shea neglected to
keep an eye on him, the business of abandoning ship on a dark
night at excessively short notice being calculated to tax the
resources of the most capable commander.
The comprador took advantage of these distractions to erase
himself from the scene. The boats were held against the side of the
steamer, while the captain took tally of the men in them, scrambling
from one boat to the other with a globe lantern swinging in his fist.
Charley Tong Sin was indubitably missing. O’Shea leaped on board
the moribund Whang Ho, which was now sobbing and gurgling
tremendously, and made a flying search of the cabins and state-
rooms. It was obvious that this elusive young Chinese had not
vanished below decks, where by now nothing but a fish could exist.
And unless Captain Michael O’Shea wished to join the fishes, it was
time for him to go.
Chagrined and anxious, he returned to his boat, and the men
frantically plied oars. A moment or two later the Whang Ho went
under with very little fuss, meeting her end with the calm of a
Chinese philosopher. The boats rocked in the waves that rolled away
from the place where she had been, and the rays of the lanterns
revealed many large and greasy bubbles.
Captain O’Shea wasted no time in sentimental regrets. The
Whang Ho was a dead issue. What vitally concerned him was the
whereabouts of that valuable passenger, Charley Tong Sin. It was
absurd to suppose that he had fallen overboard and given up the
ghost. A rascal of his kidney had as many lives as a cat. It was much
more plausible to surmise that he had unostentatiously laid hold of a
life-belt, slipped over the stern, and made for the nearest shore. The
boats moved to and fro, looking for him, but the darkness, misty and
opaque, made it hopeless to discover the head of a swimmer who by
this time might have left the water and concealed himself in the
marsh.
“I misdoubt that me policy was sound,” said Captain O’Shea to
Mr. Kittridge. “Maybe I ought to have shot him, anyhow.”
“It would ha’ been a good job,” grunted the chief engineer. “And
now he’ll streak it for this village of Wang-Li-Fu and give an alarm.”
 “Precisely that. But unless he can pick up a sampan or a fishing-
boat he will make slow headway flounderin’ through the swamps
and swimming the creeks. ’Tis up to us to beat him to it.”
Mr. Parkinson, who was in command of the other boat, was
ordered to steer alongside for consultation. It was promptly agreed
that the party should first find the mouth of the River of Ten
Thousand Evil Smells and then move up-stream without delay. It
would be slow and blundering navigation, but if three or four miles
could be traversed before daylight they might tie up to the bank and
reconnoitre within striking distance of their goal.
“I do not know what kind of a mess we will hop into,” O’Shea
told them before the boats separated. “We may have to fight our
way, thanks to that slippery divil of a comprador, and I am not
asking ye to go anywhere that I will not go meself. Some of you are
not trained to use weapons, but if ye will cut loose and blaze away
and not think too much about your own skins, we can make it
uncomfortable for a slather of Chinese. There is plenty of
ammunition, so don’t scrimp yourselves.”
The boats slid slowly into the entrance of the wide, sluggish
stream. The lanterns were extinguished. The only sound was the
cadenced thump of the thole-pins. If any of the men felt the prickly
chill of cowardice, they kept it to themselves. Now and then the
keels furrowed the mud, and when the boats stranded hard and fast,
the crews waded overboard and shoved them ahead. Thus the little
flotilla progressed until dawn flushed the eastern sky and the vapors,
streaming upward from the marshes, curled and drifted like filmy
clouds. Higher ground and the green, checkered squares of tilled
fields were discernible a short distance beyond.
The boats turned into the mouth of a tiny creek where the tall
rushes curtained them from observation. This was a favorable
halting place, and a cold breakfast was hastily eaten. O’Shea had a
poor opinion of fighting on an empty stomach. He addressed himself
with marked deference to a very neatly dressed man with iron-gray
hair who had said little during the voyage. His face was haggard and
his eyes were tired with weariness of living.
“You have seen service, sir, and ye have led drilled men,” said
O’Shea. “The cards are dealt, but from now on you can play them
better than I. I will be obliged to ye for advice.”
The cashiered officer looked grateful. This kind of recognition
had power to move him. With a diffident manner, as if his
professional opinion had long since ceased to interest any one, he
replied:
“Most Chinese villages are walled. There will be at least one gate
facing the river and two or three on the inland side. It is often
awkward to make a landing under fire from boats. I suggest we
divide our force. If you approve, Captain, I will take ten of the most
active men and disembark here. We can fetch a wide circuit of the
town, and it will not be difficult to make our way across the rice
fields and ditches. You can put the rest of them in one boat and row
up in front of the town, waiting in the stream until we are in a
position to make a rush. Then we will drive home a simultaneous
attack in front and rear.”
“Napoleon could not beat it,” heartily exclaimed O’Shea. “And if
ye shoot fast enough and kick up a terrible racket, they will think ye
are an army. What will the signal be?”
“Three rifle shots.”
“Ay, ay, Mr. Bannister. ’Tis the sensible plan that ye take
command of the army while I hoist the rear-admiral’s pennant over
the navy. We have no reserves, but many a famous victory would
have been missing from history if the lads that won them had waited
for the reserves to come up.”
The chosen ten forsook the boats and tramped off behind their
soldierly leader. A few minutes later the expedition of Captain O’Shea
got under way, his boat hugging the muddy shore and dodging
behind its ragged indentations. It was not long before a wide curve
of the river disclosed to view the tiled roofs, the crumbling brick wall,
and the towered gate-ways of a village. In front of it were several
rickety wharfs, or stagings, built of bamboo poles lashed together. At
the outer end of one of these lay a two-masted junk, her hawse-
holes painted to resemble two huge eyes. The tide had begun to
ebb, and the junk was already heeled so that her deck sloped
toward the river. This craft appeared to be deserted. No pigtailed
heads bobbed behind the immensely heavy bulwarks. If the army
officer had been a Napoleon, Captain O’Shea showed himself a
Nelson.
“Pull like blazes for the junk yonder,” he shouted to his men. “We
will pile aboard her and take cover.”
The junk was directly in front of the gate-way in the village wall,
and perhaps a hundred yards distant from it. The intervening space
was beach, a miry roadway, and a disorderly row of shanties made
of drift-wood, with a few boats hauled out for repairs. The heavy
timbers of the junk made her a nautical fortress, and the high sides
would be difficult of direct assault.
The men swung lustily at the oars, and the boat shot out into
the open river. O’Shea steered wide of the village until he could turn
and make directly for the junk. It was an admirable bit of strategy,
but wholly wasted on this sleepy, shabby Chinese village. There was
never a sign of a hostile demonstration. As an anticlimax the thing
was absurd. A crowd of men, women, and children streamed out
through the gate in the wall and stared with much excited chatter at
the foreign invaders. Apparently their behavior meant no more than
a harmless curiosity. Several garrulous old gentlemen squatted upon
fragments of timber and pulled at their bamboo pipes while they
discussed the singular visitation with the oracular demeanor of so
many owls.
The bold O’Shea grinned sheepishly. His sensations were those
of a man who beheld a heroic enterprise suddenly turned into low
comedy. He glanced at the amused faces of his followers and said:
 “’Tis not what ye might call a desperate resistance. Let us
promenade ashore and look the town over.”
They quitted their fortress and moved along the narrow, swaying
staging of bamboo, their rifles ready for use in the event of an
ambuscade. The Chinese crowd promptly retreated in noisy
confusion. O’Shea ordered a halt. After some delay, three signal
shots came down the wind from Major Bannister’s force. He was
about to attack the village from the landward side. Now the
shopkeepers and coolies scuttled madly away from O’Shea’s party to
seek shelter within the walls and discover what all this extraordinary
excitement could mean.
Behind them tramped the naval brigade into streets from which
the inhabitants were vanishing as rapidly as possible. Somewhere
near the centre of the town O’Shea and Major Bannister joined
forces. This pair of valiant leaders eyed each other with mutually
puzzled chagrin.
“We just walked in without the slightest trouble,” confessed the
army man. “What do you make of it?”
“I had the same experience,” observed O’Shea. “And I do not
know what to make of it at all. ’Twas me firm conviction that we
were prancin’ into a hornet’s nest. The information all pointed that
way. I would call it a funny kind of a surprise party.”
“The villagers have no intention of making it unpleasant for us.
They have been giving my men eggs and melons and chickens, to
keep us good-natured, I presume.”
“Well, we will find quarters and fetch our grub from the junk,
and I will buy the drinks, if ye can locate them, for the joke seems to
be on me.”
They found the village tavern, consisting of several detached
buildings set in a large court-yard. The agitated landlord kow-towed
himself almost black in the face, and in trembling accents expressed
his desire to bestow all his goods upon the warlike foreigners if only
his miserable life might be spared. He summarily ejected a few
native guests of low degree, who fled without delaying to argue the
matter. The invaders set the tavern coolies to sweeping and
scrubbing the filthy buildings and took charge of the kitchen with its
row of earthen fire-pots. There was no lack of room for men to sleep
three and four in a row upon the k’angs, or brick platforms used for
the purpose, and the ragged quilts were hung outside to air. In
short, the tavern was transformed into a camp which had no serious
discomforts.
Having taken care of his men, Captain O’Shea found leisure to
ponder over the situation, a process which left him with a headache.
He rambled unmolested from one end of the village to the other,
searching for clues that might link themselves with the Painted Joss
and the tragedy of Bill Maguire. There were two small, dilapidated
temples, one of them inhabited by a few Buddhist priests in yellow
robes. O’Shea was permitted to enter them and explore to his
heart’s content. They were nothing more than village shrines,
however, in which the perfunctory rites were held and offerings
made—such places as might have been seen in a thousand Chinese
towns. Nor did the village itself, excepting for an air of general
decay, differ from the hamlets of a dozen provinces.
“I have a harrowing suspicion that Charley Tong Sin made a
monkey of me,” ruefully sighed O’Shea, “or maybe I have been all
wrong from the start. The Chinese proposition has too many twists
in it for a white man to fathom.”
As a person of considerable confidence in his ability to master
difficulties, his self-esteem had been dealt a hard blow. His
imagination had pictured a large, stirring climax of his pilgrimage,
and here he was all adrift in a wretched little village of no
consequence whatever, the last place in the world to find the
headquarters of a secret organization so mysteriously powerful as to
cast its sinister shadow throughout China, and even across the seas.
And yet the evidence had been by no means vague and misleading.
Beginning with the fragmentary revelations of the demented sailor,
coming next to the disclosures of poor McDougal’s diary, he had
been led straight to the town of Wang-Li-Fu, on the River of Ten
Thousand Evil Smells. He had felt that the hand of destiny was
guiding him.
Returning to the tavern yard O’Shea found his men cheerfully
making friends of the villagers and accepting the situation with the
ready adaptability of true soldiers of fortune. They looked to the
leader for orders, but he had none to give them. He had been placed
in the ridiculous position of providing wages and rations for a
perfectly superfluous expeditionary force.
“Just what did you expect to turn up in this pigsty of a
settlement?” gloomily inquired Mr. Kittridge, who seemed
disappointed that he had not broken a few heads. “Whatever it was,
it fell flat.”
“It did that,” frankly admitted O’Shea. “’Tis a painful subject, Mr.
Kittridge, and we will not discuss it now. But I am not done with the
riddle of Wang-Li-Fu.”
Three days passed, and singly and in squads the invaders
ransacked the village and its suburbs, poking into shops, alleys,
dwellings, and court-yards and taking stock of the inmates thereof.
That the people were very poor and very industrious was all that one
could say of them. And they were no more to be suspected of
plotting deeds of violence than so many rabbits. Doggedly
persistent, unwilling to confess himself beaten, O’Shea shifted his
quest to the open country for miles outside of Wang-Li-Fu. It was a
region of green fields gridironed with ditches and rutted paths, and
dotted with toilers in blue cotton blouses and straw hats, who tilled
their crops from dawn to dark.
It was obviously useless to extend the investigation any
considerable distance away from this region. If the secret was not to
be unearthed in the vicinity of Wang-Li-Fu, then his conclusions had
been all wrong. The villagers assured him that this was, in truth,
none other than Wang-Li-Fu, and the baffled, perplexed O’Shea
could not let go of the opinion that the goal was somewhere near at
hand. Otherwise, why all the elaborate stratagems in Shanghai to
thwart his voyage to the River of Ten Thousand Evil Smells?
He had imagined himself attacking a stronghold of some sort, a
headquarters of desperate criminals who must be wiped out. But if
that slippery comprador Charley Tong Sin had carried a warning to
the men of the Painted Joss, he must have fled elsewhere than to
this commonplace, harmless village. At any rate, it seemed absurd to
tarry much longer in Wang-Li-Fu with a force of armed retainers.
At the end of a fortnight, O’Shea was of the opinion that his loyal
legion had better seek to mend its fortunes in some other quarter.
He was ashamed to look them in the face. The fiasco cut him to the
quick. He had been as mad as poor Bill Maguire. In future he would
stick to his trade as a shipmaster.
Meanwhile, the malarial poison of the marshes found its way into
his blood. He failed to realize that he was ill, and paid no attention
to the little flashes of fever that came by night and the creeping,
chilly feeling that troubled him in the morning.
There came a day when he was unable to rise from the brick
sleeping-platform. The fever increased, suddenly, violently. It caught
him unprepared. His plan of retreat had not been announced, and
now he was incapable of leadership. His mind alternated between
delirium and stupor. When he talked it was of many inconsequential
things. One might have said that the evil spirit of the Painted Joss
had laid its spell of misfortune upon him. In the court-yard of the
tavern his lieutenants held a conference.
“Can anybody make head or tail of this infernal situation?”
gloomily inquired Mr. Kittridge. “What in hades are we going to do
about it?”
“Try to pull Captain O’Shea through this fever before we think of
anything else,” stoutly affirmed Mr. Parkinson. “We jammed into this
crazy voyage with our eyes shut. With all of us it was anything to get
clear of Shanghai. And it’s useless business to sit and growl about it
as hard luck. What do you say, Major Bannister?”
 The army man smiled at sight of their discouraged countenances
and quietly answered:
“What else can we chaps expect but hard luck? Really, I should
be surprised to find anything else. I can tell you one thing,
gentlemen. I have campaigned in the tropics, and I know something
about this swamp fever. We had best get out of here and take
Captain O’Shea with us. If we don’t, he will die as sure as sunrise,
and the rest of us will be down with it before long. It caught him
first because he was fagged with worry.”
“We agree with you there,” said Mr. Parkinson. “But we seem to
have overlooked a line of retreat. That was the Irish of it, I suppose.
If we go down river in our two boats we’ll have to work ’em out to
sea over those nasty shoals and then run the chance of being picked
up adrift. We might get away with it, but it would kill a man as sick
as O’Shea.”
“Why not go up-river?” suggested Major Bannister. “By means of
a few words of Chinese and a great many gestures I have extracted
from the village head-men the information that there is a European
mission station about a hundred and fifty miles northwest of here.
We can make part of the journey by boat and then hike overland.
With a litter and coolies to carry it, we may be able to take Captain
O’Shea through alive. It’s better than letting him die in this pest-
hole.”
“That’s the most sensible speech I’ve heard since we signed on,”
grunted Mr. Kittridge. “And you can pull out of this rotten Wang-Li-Fu
not a minute too soon to please me.”
The village head-men were summoned, and these venerable
worthies declared themselves anxious to aid the sick leader of the
foreign soldiers. He had played with their children, paid the
shopkeepers their prices without dispute, and sat with the old men
in the tea-houses. Nor had his armed force committed any abuses,
although they held the village at their mercy. It was wisdom to try to
carry Captain O’Shea to his own people. The village would gladly
furnish a guide and plenty of coolies, a covered litter, and a small
house-boat in which the sick man could be made comfortable.
The evacuation of Wang-Li-Fu was a dismal business. The
adventurers were oppressed by a sense of failure and
discouragement. Their enterprise had fizzled out like a dampened
match. This final act was inglorious. Their plight was worse than
when they had been stranded as beach-combers in Shanghai. They
carried Captain O’Shea to a sampan, or flat-bottomed boat, with a
tiny cabin of bamboo and matting, which could be towed against the
sluggish current of the river. The men disposed themselves in the
two boats saved from the Whang Ho steamer, and a squad of half-
naked coolies strung themselves along a towing-rope to help track
the sampan up-stream.
The sick man lay stretched upon his quilts and showed little
interest in the slow progress of the flotilla. Between spells of heavy
drowsiness he watched the slimy shore and fringing marsh slide
past. Through the first day the wind was cool and the air bright, and
the boats trailed up-river until after nightfall before they were pulled
into the bank to moor. As the part of caution, no fires were made
and conversation was hushed. The foreigners had an uncomfortable
suspicion that this might be hostile territory, although they had
discovered nothing to warrant the conjecture. But O’Shea had been
babbling about the Painted Joss while flighty with fever, and Charley
Tong Sin was still unaccounted for.
Between midnight and morning the sick man came out of his
uneasy dreams. As it seemed to him, he was clear-headed, his
senses alert, his judgment normal. Just why he should be cooped up
in this native boat was a bit difficult to comprehend, but why try to
understand it? There was only one problem of real importance. And
now was the time to solve it. O’Shea laughed to think what a stupid,
blundering fool he had been to recruit an armed expedition and
come clattering into this corner of China with so much fuss and
noise.
 If a man wanted to find the Painted Joss, all he had to do was
listen to the friendly, familiar voices that whispered in his ears.
O’Shea could hear them now. He accepted them as a matter of
course. His eyes were very bright as he pulled on his shoes and
fumbled for the revolver in its holster under the pillow. Curiously
enough, he was no longer conscious of great physical weakness. It
was tremendously urgent that he should go to find the Painted Joss
without a moment’s delay. His men would not understand if he
should tell them about the friendly voices that were offering to show
him the way. They might try to restrain him. He must leave the boat
quietly, unobserved.
Crawling from beneath the matting curtain, he gained the river
bank. His knees were exceedingly shaky and his hands trembled
uncertainly, but he was confident that he had found the trail of the
Painted Joss and that his vigor would soon return. Charley Tong Sin
outwit him? Nonsense! O’Shea would have been startled beyond
measure to know that he was wandering off in delirium. He would
have taken a shot at any one rash enough to tell him so.
Undetected he moved along the shore, silent as a red Indian,
and was presently lost in the darkness. It was muddy walking, and
he turned into the tall marsh grass, where a carpet of dead
vegetation made firmer footing. Frequently he was compelled to halt
and regain his labored breath, but his purpose was unwavering. The
voices drove him on. He had no sense of fear. After some time his
erratic progress led him back to the river. There he stumbled over a
log and sat down to wait for daybreak, which had begun to flush the
sky.
His head throbbed as though hammers were pounding it and
waves of blurring dizziness troubled him. What was more
disquieting, the guiding voices had ceased to talk to him. He felt
crushing disappointment and sadness. His eyes filled with tears.
Dawn found him seated dejectedly with his back propped
against the log, his head drooping, while he stared at the muddy
river. Here he would wait on the chance that his friends might find
him. As the day brightened, his aimless vision was caught by
something which powerfully awakened his weary, befogged
perceptions. It acted as a stimulant of tremendous force. Sitting bolt
upright he gazed at a footprint, cleanly outlined, which had become
sun-dried and hardened in a stratum of clay.
It had been made by a leather sole and heel. The outline was
pointed and narrow. Into O’Shea’s quickened memory there flashed
the picture of Charley Tong Sin stretched upon the cabin floor of the
Whang Ho steamer, his patent-leather shoes waving gently as he
went to sleep under the soporific influence of a knock-out blow. He
felt absolutely certain that this particular print had been left by the
fashionable footgear of the vanished comprador. The voices had
guided him aright. It was here that Charley Tong Sin had come
ashore after making his way up the River of Ten Thousand Evil
Smells in some kind of a native boat.
There was one chance in a million that O’Shea should have
halted to wait in this precise spot where his eyes might see the
thing. He dragged himself to his feet and scanned the melancholy
landscape. There were no villages in sight; only the marsh and fields
and a vast mound of débris to mark the place where once had stood
a city. Even the walls surrounding it had been levelled. It was
scarcely more than a wide-spread excrescence of broken brick and
tiling partly overgrown with vegetation. The landscape could have
held no more desolate reminder of the wreckage left in the wake of
the Tai Ping rebels.
It was plausible to surmise that this was the real Wang-Li-Fu,
the city which O’Shea had set out to find. The squalid village much
lower down the river might have been founded by refugees who
gave the same name to their new abode. And the villagers had been
too ignorant to explain the blunder. To them there was only one
Wang-Li-Fu. How Charley Tong Sin must have laughed at leaving
O’Shea and his men to waste themselves in a chase that led
nowhere.
 It was a pallid, unshaven, tottering ghost of Captain Michael
O’Shea that mustered strength to walk very slowly in the direction of
the ruined city. Once he paused and became irresolute, but a little
way beyond he found the imprint of a narrow shoe of European
workmanship on the soft bank of a ditch. His stumbling steps led
him, as by an unerring divination, toward the highest part of the
great mound of débris where tall trees grew from the crumbling
masonry. His painful advance became less difficult when he found a
path from which the obstructions had been removed.
Presently he stood looking across a cleared space in the midst of
the ruins, invisible from river or highway. In it were several small
buildings and one much larger. The timbers set into its walls were
carved and gilded, the curving roof of dull red tile. There was no
living thing in sight. This isolated community was so situated that it
was wholly concealed from strangers, and the natives of the region
were apt to shun the blasted city as haunted by demons. No
watchers were posted to guard against intrusion.
O’Shea crossed the open space and made for the large building,
which had the aspect of a temple. Unhesitatingly he approached the
massive wooden doors and found them ajar. He walked like a man in
a trance, muttering to himself. Passing within, he entered a sort of
anteroom partitioned by means of screens wonderfully embroidered.
The stone pavement rang to the tread of his heels. The place
echoed with emptiness. He pressed on and came into a room of
greater extent. Its corners were lost in shadow. Rows of pillars
supported the dusky rafters upon which gilded dragons seemed to
writhe. The windows were small and set close to the roof and the
light of early morning had not dispelled the gloom.
In the centre of the floor was an altar. Behind it towered an
image of Buddha, and yet it was unlike the images of the bland and
contemplative Buddha commonly to be found in the temples of the
East. It was a monstrous thing. Only an artist with an inspiration
from the devil could have so handled tools as to make those wooden
features seem to lust after all abominable wickedness. The color of
this seated statue was crimson. Amid the shadows it glowed like fire
or blood. On the breast, above the folded arms, stood out in broad,
black strokes a Chinese symbol or character which O’Shea
recognized with a sensation of creeping repugnance.
“The Painted Joss!” he gasped.
His attention was so strongly caught and held by this malevolent
image that for the moment he had eyes for nothing else. Presently,
however, he became aware that another figure confronted him, a
living presence. It was a man sitting in a massive chair of teak-wood,
by the side of the Buddha. The bulk of him was enormous. He was
both fat and mighty of frame, and not even the towering amplitude
of the image could dwarf his proportions and belittle the impression
he conveyed. His face was broad and heavy-jowled, the mouth
sensual and cruel. With folded arms he sat and gazed at the foreign
intruder. This unflinching, scornful immobility had a certain
distinction. He believed that he must instantly die at the hands of
this European with the white, savage face and the blazing eyes who
covered him with a revolver. It was futile to cry out and summon
help. As is customary with Chinese in positions of authority, this
high-priest of iniquity had gone to the temple to have audience with
his servitors very early in the morning. They had not yet joined him
and O’Shea was quick to read his own advantage.
It was right and just that he should slay this huge man in the
crimson robe who ruled the temple of the Painted Joss. He had
come ten thousand miles to be judge and executioner. He was ready
to kill and be killed in his turn. But the revolver was strangely heavy
and it wavered so that he was unable to hold it at arm’s length. A
haze bothered his vision and he could not brush it from his eyes.
Something was the matter with his knees. They were giving way.
With an incoherent exclamation, O’Shea fell unconscious upon the
stone flagging and the revolver clattered from his limp hand. He had
paid the price of exertion beyond his strength.
When his senses returned there was in his mind only the
dimmest recollection of how he came to be in this dreadful place.
The vagaries of fever no longer possessed him. Clear-headed but
wretchedly weak and nerveless, he gazed about him and discovered
that he was alone in the unholy temple. The shadows were not so
heavy on the pillars, the gilded rafters, and the marble altar. The
crimson image of the seated Buddha loomed flamboyant and
portentous and the Chinese symbol painted on its breast was boldly
outlined.
There was no way of escape. The building was a most effectual
prison. His revolver had been taken from him. He could not even
fight and die like a man. The fact was that this desperate extremity
lacked the proper sense of reality. It was so contrary to reason and
he had such shadowy, confused ideas of what had preceded, that
this was more like nightmare or delirium. And it seemed impossible
that he should not presently find himself awake.
What most tenaciously persisted in his memory was the image of
the huge man in the teak-wood chair. He was a vision which could
not be denied. Such a one as he had power to sway the wills of
others to his desires, to create and direct great enterprises and send
his influence afar, but never for good. If he ordered murder done in
distant places his secret edicts would be obeyed, nor would his
agent dare to thwart him. If there was such an organization as
O’Shea had assumed, then he had stood face to face with the
dominant personality, the compelling force from which radiated
infernal activities.
“I saw him, whether I am meself or somebody else,” the
prisoner muttered with a groan. “And he will come back and the
brand will be chopped into me, same as was done to poor Bill
Maguire. ’Tis a tough finish, if all this is really true. My God, I wish I
knew what had happened to me. Yesterday I was going up-river
with me men, and now——”
He struggled to his feet. A supreme effort of will conquered
physical weakness. A man condemned to die is capable of forgetting
bodily ills. Just then a young man appeared from the direction of the
door-way. He wore native garments, but O’Shea recognized him. It
was Charley Tong Sin, whose smile was unpleasant. In his hand was
O’Shea’s revolver, which he was careful to hold ready for use. The
jaunty, affable manner of the comprador had returned. He appeared
very well satisfied with himself as he exclaimed, by way of greeting.
“It is an unexpected pleasure, you bet, Captain O’Shea. I have
waited till you were gone from Wang-Li-Fu. It was reported that you
were very sick and went up the river yesterday with your men. You
decided to come and see us, to visit the Painted Joss? You wished to
make some trouble?”
“’Tis the last day I will make trouble for any one, by the looks of
things,” replied O’Shea. “You win, Charley.”
“You are a smart man,” grinned the other. “But you had too
much curiosity. I am a good fellow. I will tell you what you want to
know. You will not give it away. They are getting ready to cut your
visit pretty short.”
There was the chatter of voices somewhere outside and the
brazen mutter of a gong. O’Shea kept silence. He was not as
resigned to his fate as Charley Tong Sin inferred. He was watching
every motion of the gloating young man and his eyes measured the
distance between them.
“You will feel better if you know,” tauntingly cried the Chinese.
“You have seen the Painted Joss. You have seen a man sitting beside
it, the great and terrible Chung himself, the ruler of the Pih-lien-Kiao,
the Sect of the Fatal Obligation.”
“Much obliged, Charley,” grimly interrupted O’Shea. “Tell me
some more. I am sorry I could not have words with the terrible
Chung. And the brand that ye chop into people, your trademark?”
“It is the mark that means The Dreadful Messenger of Chung. It
is a favor to tell you, Captain O’Shea. No other foreigner, no Chinese
except the servants of Chung, have heard it spoken. But you will not
speak it anywhere.”
 “There’s more that I want to know,” said O’Shea, “though
precious little good the information will do me.”
“Ha! Why did you not have so much sense before and mind your
own business?”
It was absurd to carry on such a dialogue as this, as O’Shea
perceived, but Charley Tong Sin was enjoying this session with the
rash shipmaster who had formerly held the upper hand. Before the
victim could be subjected to further taunts he heard the massive
doors opened and other sounds to indicate that bars were sliding
into place to fasten them on the inside. The huge man in the
crimson robe, the great and terrible Chung, lumbered into view and
seated himself in the chair of teak-wood. Charley Tong Sin humbly
bowed several times. The personage beckoned the twain nearer and
spoke briefly. He desired to conduct a cross-examination of his own
with the comprador as interpreter.
“He wishes to know why you have come to this place?” was the
first question addressed to O’Shea.
“Because ye butchered a friend of mine, a red-headed sailor by
the name of Jim Eldridge,” was the unflinching reply. “He told me
about your dirty devilment as well as he could, and I saw what ye
did to him.”
The huge man showed signs of consternation when this was
conveyed to him. He uttered a bellowing interrogation.
“He is not alive? You have talked with his ghost?” shrilly
demanded Charley Tong Sin.
“’Twas him that sent me here,” declared O’Shea. “Ye can impart
it to the big ugly mug yonder that I have had visits from the ghost of
the red-headed sailor that he killed and branded.”
With an excited, heedless gesture, Charley Tong Sin raised the
revolver. He had been long accustomed to wearing European
clothes, and the flowing sleeves of his Chinese outer garment
impeded his motions. A fold of the silk fabric fell over the butt of the
weapon, and he tried to brush it aside with his left hand. This other
sleeve was caught and held for a moment by the sharp firing-pin of
the cocked hammer.
This trifling mishap, gave O’Shea a desperate opportunity. With a
flash of his normal agility he leaped across the intervening space.
The comprador strove frantically to free the weapon, but only
entangled it the more. The episode was closed before the crimson-
robed personage could play a part. O’Shea’s shoulder rammed
Charley Tong Sin and sent him sprawling, and the revolver was
instantly wrested from his grasp.
“The doors are locked,” panted O’Shea, “and before your men
break in, I will send the both of ye to hell. Sit where you are, ye
terrible Chung. You overplayed your game, Charley.”
The comprador seemed to shrink within his clothes. His mouth
hung open and his face was ashen. He was eager to clutch at any
straw which might give him the chance of life. Shrinking from the
scowling presence in the chair, he began to talk a sing-song babble
of words that tumbled over each other.
“I will help you get away alive if you do not kill me. Captain
O’Shea, I will explain about Jim Eldridge; I will not lie to you. All the
secrets I will tell you. There was a steamer, the Tai Yan, and she
came over the bar from the sea in a big storm, at the time of a
flood. It was do this or go to the bottom because the engines had
broke. A boat with sailors rowed up the river. They were foolish men
who believed the stories that gold and silver treasure was hidden in
the ruins of this old Wang-Li-Fu. And they found this temple, and
they knew too much.
“All but two of the men were able to run quick to the river, but
Eldridge and one named McDougal ran into this place, trying to hide.
They ran into the temple before they were captured. There was a
little building, but now it is ashes and much sticks of burnt wood. In
that building those two men were locked to be killed next day. The
red-headed man was a demon, I tell you. Walls could not hold him.
In the night he set fire to the building, and it was a great blaze. But
he was caught and punished.”
“Ye left him for dead, and he came to,” growled O’Shea. “And so
McDougal got away!”
“I can tell you more secrets,” wailed Charley Tong Sin, but his
services as an informer were suddenly cut short. The huge man in
the chair had raised his voice in a tremendous call for help to his
followers without. Otherwise he had sat composed, glaring at
O’Shea. It was his hand that slew Charley Tong Sin as a traitor. He
was on his feet, the heavy chair raised aloft. He swung it with
amazing ease. It was no longer a massive article of furniture, but a
missile in the hands of a man of gigantic strength. His movements
were not clumsy.
The chair flew through the air. O’Shea dodged, but Charley Tong
Sin flung up his arms, taken unawares. The impact would have
brained an ox. The whirling mass of teak smote the terrified
comprador on the head and chest and he crumpled to the
pavement. He was as dead as though he had been caught beneath
the hammer of a pile-driver. The tableau was an extraordinary one.
O’Shea stood staring at the broken body of the young Chinese. The
man in the crimson robe stirred not from his tracks. Implacable,
unafraid, he had executed the last sentence of The Sect of the Fatal
Obligation.
The people outside were clamoring at the doors, and O’Shea
heard the thud and crash of some kind of an improvised battering-
ram. He sighed and found the thought of death at their hands very
bitter. But he would not go alone. He faced the great and terrible
Chung and slowly raised the revolver.
The arch-assassin bade him wait with a gesture so imperious, so
mandatory, that O’Shea hesitated. The bearing of the man held
some large significance. His dark, evil countenance expressed rather
sadness than wrath. He slid a hand into the folds of his robe and
raised the hand to his mouth. Whatever it was that he swallowed
wrought its work with swift and deadly virulence. Swaying like a tree
about to fall, he strode to the marble altar and fell across it with his
head buried in his arms. In this posture he died, in front of the
image of the glowing Buddha, whose graven lineaments seemed to
express the unholy ambitions and emotions of his own soul.
O’Shea managed to walk to a corner of the temple and slumped
down upon a marble bench where the Painted Joss cast its deepest
shadow. His strength had ebbed again. Listlessly, almost inattentive,
he heard the assault upon the doors renewed and the splintering of
plank. When the Chinese mob came tumbling in he could try to
shoot straight and hit a few of them, and then they would close in
on him. It was the end of the game.
A few minutes and the servitors of Chung came jostling and
shouting through the anteroom. Then they halted abruptly. Their
noise was hushed. The light that fell from the windows near the roof
showed them the lifeless figure in the crimson robe, doubled across
the marble altar. In the foreground lay the battered body of Charley
Tong Sin, but they had eyes only for the tragedy of the altar. They
stood dumfounded, like men in the presence of something
incredible.
At length the boldest shuffled forward. The others followed
timidly. They appeared terrified in the extreme. It was as though
they had believed their master to be invulnerable. And he was dead.
Possibly they conjectured that he had been slain by an agency more
than mortal. The group of Chinese clustered about the altar,
whispering, regarding the body of Chung. Apparently they had not
bethought themselves of the foreigner who was held a prisoner in
the temple.
O’Shea rose in his shadowy corner and moved wearily past the
Painted Joss. It was better to have the thing finished. He came upon
the Chinese like an apparition. Their wits were so fuddled that the
sight of him had the effect of another shock. If he had been
powerful enough to slay the mighty Chung, then the demons were
his allies. Perceiving their dazed condition, he forebore to shoot, and
advanced abreast of the altar. The path to the door-way was clear,
but he had not the strength to make a run for it. The hope of life,
miraculously restored to him, was in the possibility that they might
stand and gaze at him a little longer.
He had walked a half-dozen steps farther when one of the crowd
yelled. The spell was broken. They raced after him like wolves. He
turned and steadied himself and pulled trigger until the revolver was
empty. The onset was checked and thrown into bloody confusion.
O’Shea had summarily convinced them that whether or not the
demons were in league with him, the devil was in this ready weapon
of his.
They were no longer massed between him and the exit, and for
the moment the advantage undeniably belonged to this mysterious,
devastating foreigner.
He stumbled over the broken timbers of the doors and was in
the blessed daylight, the temple behind him. He would be overtaken
ere he could flee the ruined city, but he reloaded the revolver as he
followed the path at a staggering trot. The mob poured out of the
temple, yelping in high-keyed chorus. As a foot-racer the hapless
Captain Michael O’Shea was in excessively poor condition. In fact, it
promised to be the easiest kind of a matter to overtake him and
leisurely pelt him to death with bricks as soon as he should have
expended his ammunition.
He swerved from the rough path and crawled to the top of a low
ridge of débris. Standing erect for a moment, he pitched forward
and fell against a bit of wall. His figure had been outlined against the
sky, and it was discerned in a fleeting glimpse by a scattered band of
men in khaki and linen clothes who were tramping the marsh. They
raised a shout and rushed toward the ruined city, converging until
the force was mobilized within a short distance of the prostrate
O’Shea.
The Chinese mob, pursuing full-tilt, found itself confronting a
score and more of rifles which enthusiastically opened fire until the
air hummed with bullets. There was a hasty, unanimous retreat of
the followers of Chung to the temple and the adjacent buildings.
Major Bannister halted to bend over O’Shea and say:
“We thought you were drowned or bogged in the marsh. What
sort of a rumpus is this?”
“The Painted Joss,” murmured O’Shea. “I found it. Don’t bother
with me. Go to it and clean out the place.”
The adventurers, at last earning their wages, proceeded to make
things most unpleasant for the household of Chung. The resistance
was brief, and those who were not penned within the temple fled in
panic and sought cover in the marsh. They were taken by surprise,
for the community had found the visit of Captain O’Shea sufficient to
engage its attention. To him returned Major Bannister, hot and dusty,
his cheek bleeding from the cut of a Chinese sword, and smilingly
announced:
“Bully good fun while it lasted. What shall I do with the devils we
cornered? Take them out and shoot them?”
“No. The boss of the works is dead. And I have a notion that
The Sect of the Fatal Obligation died with him. Lug me to the
temple, if ye please. I’m all in, but ’tis my wish to see the whole
wicked business go up in smoke.”
Before the torch was applied, that experienced man of war,
Major Bannister, suggested that he had never seen a more promising
place in which to poke about for loot. The search amounted to
nothing until it occurred to the major to pull the Painted Joss from
off its pedestal. After much heaving and prying the great image fell
crashing to the pavement of the temple. Investigation revealed that
underneath it were several compartments accessible by means of
cunningly fitted panels. Many papers or documents were found,
wrapped in silk, and it was assumed that these were the records of
the black deeds of Chung and his organized murderers. They were
thrown aside, to be bundled together and taken to the boats.
 It was the astute Major Bannister who smashed the bottom of
one of these compartments with a rifle-butt and rammed his hand
through the splintered hole. His groping fingers came in contact with
closely packed rows of metal bars. In this manner was discovered
the wealth of the temple, the blood-money stored and treasured by
the infamous Chung, the price of many assassinations.
The gold was in stamped ingots, the silver in the lumps or
“shoes” of the clumsy Chinese currency, and there were baskets of
English sovereigns, Mexican dollars, and a variety of the coinages
which pass over the counters of the money-changers of the Orient.
Murder as a business had paid well. The Sect of the Fatal Obligation
was a flourishing concern. The loot belonged to those who found it.
They were troubled by no scruples respecting the heirs of the
departed Chung, nor did they consider it their duty to surrender the
spoils to the Chinese government.
That night a conflagration reddened the ruins of the dead city of
Wang-Li-Fu. It was the pyre of the Painted Joss. And when the little
flotilla again moved up-river early next morning, a cloud of smoke
rose lazily in the still air. Captain Michael O’Shea was still alive, which
was rather surprising, for he had passed through experiences
extremely disturbing to a sick man. There was tonic, however, in the
fact that he had redeemed his failure, the expedition was no longer
a sorry jest, and the account of Bill Maguire had been squared.
He slept with tremendous earnestness through a night and a
day, and when he awoke it was to roar for food and to display the
peevish temper of a genuine convalescent. When off duty his
comrades became absorbed in the odd occupation of arranging piles
of gold bars, silver “shoes,” and minted coins on the deck of the little
house-boat, like children playing with blocks. They smiled a great
deal and talked to themselves. Captain O’Shea looked on with an air
of fatherly interest. After all, this happy family of his had made a
prosperous voyage of it. Dreams of rehabilitation cheered these
broken wanderers. They would go home. No more for them the
misery, the heartache, the humiliation of the tropical tramp. Their
riches might slip through their fingers, but they would make the
most of golden opportunity. Like poor McDougal, they had thrown all
regrets away.
“’Tis share and share alike,” said O’Shea, “but there is a red-
headed sailor-man at anchor on a farm in Maine and I think he has a
wife somewheres. With your permission we will deal him a share of
the plunder. ’Twas poor Bill Maguire that gave us the tip.”
Unmindful of labor and hardship, this contented company slowly
journeyed to the head of navigation on the River of Ten Thousand
Evil Smells and then trudged overland while O’Shea rode in a
covered chair and sang old sea-chanties in a mellow voice. When, at
length, the English mission station was reached it was stretching the
truth to call him an invalid. The senior missionary, a gentle, very
wise old man who had lived for thirty years in the back country,
heard the tale told by these tanned, ragged travellers and was
horrified that such things should have existed. But he had news for
them, and it was thus that he supplied a missing fragment of the
puzzle of Bill Maguire:
“The man came here and we took care of him. But there was no
finding out how he had been so frightfully hurt. He was dumb and
stupid. Later I met a native boatman who had found him on the
river-bank near Wang-Li-Fu. Evidently he had been thrown into the
water as an easy way to get rid of the body. Reviving a little, he
splashed his way ashore or the tide left him there. He stayed with us
until he was fairly strong and one morning he was gone.”
“And did he set the house afire?” inquired O’Shea.
“Why, there were two accidental fires in the compound at that
time, but we laid it to the carelessness of the kitchen coolies,” was
the innocent reply.
“It was Bill Maguire, all right,” declared O’Shea. “Now, will ye be
good enough to look over the Chinese documents we found hid
away under the Painted Joss?”
 The missionary pored over the papers for several hours. And his
painstaking translation revealed all that O’Shea cared to know
concerning the operations of The Sect of the Fatal Obligation. It had
worked in secret to remove enemies for a price. If a merchant
wished a business rival obliterated, if an official found others in his
way, if it was advantageous to create a vacancy in some other
quarter, the murder guild directed by the departed Chung would
transact the affair, smoothly, without bungling. And those who knew
and would have disclosed the secret were frightened into silence by
the sight of the brand that was called The Dreadful Messenger of
Chung.
“It will interest you to learn, as an American, Captain O’Shea,”
said the missionary, “that among these documents is a list of
persons proscribed or sentenced to be slain. The most conspicuous
name I find to be that of the Chinese ambassador to the United
States, His Excellency Hao Su Ting. It is probable that this terrible
fate would have awaited him upon his return to his own country.”
“They potted his brother,” exclaimed O’Shea. “And he was sick
with fear of the thing, for I talked it over with him meself. Well, he
can thank Bill Maguire for letting him die in his bed when his proper
time comes.”
Three weeks later Captain O’Shea sat at his ease upon the
piazza of the Grand Hotel, that overlooks Yokohama Bay. He was
thinner than when he had put to sea in the Whang Ho steamer, but
he appeared to find the game of life quite worth while. It was his
pleasure to enjoy the tame diversions of a tourist before boarding a
mail-boat for the long run home to San Francisco. He smiled as he
reread a letter written in the crabbed fist of that zealous
agriculturist, Johnny Kent, who had this to say:

Dear Captain Mike:

The Lord only knows what trouble you’ll be in when this gets
to China. My advice is to quit it and come home. I’m worried
about you. Bill Maguire has rounded to, understand? His busted
main hatch sort of mended itself by degrees. He had symptoms
before you left, and you ought to have waited, but I suppose
you can’t help being young and Irish.
He was terrible melancholy at first, and he ain’t real spry
yet. I found his wife and little girl for him in Baltimore, and
made them come on here. You guessed right about the wax doll.
I bought the darndest, biggest one I could find. Bill feels that
the family is living on my charity, and being morbid and down-
hearted, he frets a whole lot about being broke and stranded.
He’ll be no good to go to sea again. It gives him the shivers to
talk about it. I don’t need him as a farm-hand in the winter, and
as for having his wife as a steady house-keeper, I’m fussy and
set in my ways.
Bill got up against an awful bad combination in China. I
won’t tell you where it was, for I don’t want you to find it.
Maybe you’ll run across a man named McDougal out there. He
was with Bill when they got in trouble. Bill saw a chance to get
away in the night, but he stood the crowd off somehow to give
McDougal leeway to join him. And this McDougal lit out with
never a thought for Bill. There was something wrong with
McDougal, as I figure it out. Maybe he was a good man, but
here was one time when he fell down on his job. None of us say
much about it, Captain Mike, but we all pray we won’t get
caught that way. You know what I mean. We’re afraid there may
be a weak spot in us that we don’t know is there until we have
to face the music. Anyhow, as I gather from Bill, McDougal was
a quitter. If I know anything about men, he has wished a
hundred times since that he had stayed to take his medicine
with Bill. We would a heap sight rather see you come home alive
than to go monkeying with the Painted Joss. Nothing much has
happened except a dry spell in August and corn and potatoes
set back. Hens are laying well.
Your friend,
J. Kent.
 Captain O’Shea chuckled and then became thoughtful. Paddy
Blake and McDougal. Charley Tong Sin and the wreck of the Whang
Ho. Wang-Li-Fu and the terrible Chung. Much can happen within the
space of a few weeks to a man that will seek the long trail. Presently
he took from his leather bill-book several slips of paper which he had
received from the Yokohama Specie Bank in exchange for his gold
bars and silver “shoes.” After making sundry calculations with a
pencil, he said to himself:
“The share of Jim Eldridge, alias Bill Maguire, is nine thousand
eight hundred and sixty-two dollars and eleven cents, and ’tis here
all ship-shape in two drafts on New York. My piece of the loot is the
same. But the red-headed sailorman will never be the lad he was,
and he should not be worried by the lack of money to live on. And
could any money pay for what he went through? ’Tis easy to know
what I should do. I will not take a cent of the plunder. My share I
will give to Bill, and with his bit of it he will be comfortably fixed.”
An expression of boyish satisfaction brightened his resolute
features as he added:
“A man would be ashamed to take money for such a pleasant
vacation as this one has been. Now, I will send a cable message to
Bill Maguire and it will cheer him a lot. His account is squared. And I
think I have put a crimp in The Sect of the Fatal Obligation.”
THE END
 Transcriber’s Notes
Punctuation, hyphenation, and spelling were made
consistent when a predominant preference was found in the
original book; otherwise they were not changed.
Simple typographical errors were corrected; unbalanced
quotation marks were remedied when the change was
obvious, and otherwise left unbalanced.
Page 319: “slung-shot” is a maritime tool.