PDF Probability Theory and Statistical Inference Empirical Modeling With Observational Data 2Nd Edition Spanos A Ebook Full Chapter

Probability theory and statistical
inference empirical modeling with

observational data 2nd Edition Spanos
A
Visit to download the full and correct content document:
https://textbookfull.com/product/probability-theory-and-statistical-inference-empirical-
modeling-with-observational-data-2nd-edition-spanos-a/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Probability Theory And Statistical Inference Empirical

Modeling With Observational Data Aris Spanos
https://textbookfull.com/product/probability-theory-and-
statistical-inference-empirical-modeling-with-observational-data-
aris-spanos/
Probably Not: Future Prediction Using Probability And

Statistical Inference Lawrence N. Dworsky
https://textbookfull.com/product/probably-not-future-prediction-
using-probability-and-statistical-inference-lawrence-n-dworsky/
Exact Statistical Inference for Categorical Data 1st

Edition Shan
https://textbookfull.com/product/exact-statistical-inference-for-
categorical-data-1st-edition-shan/
Applied statistical inference with MINITAB Second

Edition Lesik
https://textbookfull.com/product/applied-statistical-inference-
with-minitab-second-edition-lesik/
Theory of stochastic objects probability stochastic
processes and inference 1st Edition Micheas
https://textbookfull.com/product/theory-of-stochastic-objects-
probability-stochastic-processes-and-inference-1st-edition-
micheas/
Theory of stochastic objects: probability, stochastic

processes, and inference First Edition Micheas
https://textbookfull.com/product/theory-of-stochastic-objects-
probability-stochastic-processes-and-inference-first-edition-
micheas/
Statistical independence in probability analysis number

theory Kac
https://textbookfull.com/product/statistical-independence-in-
probability-analysis-number-theory-kac/
Computer Age Statistical Inference Algorithms Evidence

and Data Science 1st Edition Bradley Efron
https://textbookfull.com/product/computer-age-statistical-
inference-algorithms-evidence-and-data-science-1st-edition-
bradley-efron/
Computer Age Statistical Inference Algorithms Evidence

and Data Science 1st Edition Bradley Efron
https://textbookfull.com/product/computer-age-statistical-
inference-algorithms-evidence-and-data-science-1st-edition-
bradley-efron-2/
Probability Theory and Statistical
Inference
Doubt over the trustworthiness of published empirical results is not unwarranted

and is often a result of statistical misspecification: invalid probabilistic assump-
tions imposed on data. Now in its second edition, this bestselling textbook offers
a comprehensive course in empirical research methods, teaching the probabilistic
and statistical foundations that enable the specification and validation of statistical
models, providing the basis for an informed implementation of statistical proce-
dure to secure the trustworthiness of evidence. Each chapter has been thoroughly
updated, accounting for developments in the field and the author’s own research.
The comprehensive scope of the textbook has been expanded by the addition of
a new chapter on the Linear Regression and related statistical models. This new
edition is now more accessible to students of disciplines beyond economics and
includes more pedagogical features, with an increased number of examples as well
as review questions and exercises at the end of each chapter.
ARIS SPANOS is Wilson E. Schmidt Professor of Economics at Virginia Poly-

technic Institute and State University. He is the author of Statistical Foundations
of Econometric Modelling (Cambridge, 1986) and, with D. G. Mayo, Error and
Inference: Recent Exchanges on Experimental Reasoning, Reliability, and the
Objectivity and Rationality of Science (Cambridge, 2010).
Probability Theory
and Statistical
Inference
Empirical Modeling with
Observational Data
Second Edition
Aris Spanos
Virginia Tech
(Virginia Polytechnic Institute & State University)
University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9781107185142
DOI: 10.1017/9781316882825

c Aris Spanos 2019
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 1999
Third printing 2007
Second edition 2019
Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Spanos, Aris, 1952– author.
Title: Probability theory and statistical inference : empirical modelling
with observational data / Aris Spanos (Virginia College of Technology).
Description: Cambridge ; New York, NY : Cambridge University Press, 2019. |
Includes bibliographical references and index.
Identifiers: LCCN 2019008498 (print) | LCCN 2019016182 (ebook) | ISBN 9781107185142 | ISBN
Subjects: LCSH: Probabilities – Textbooks. | Mathematical
statistics – Textbooks.
Classification: LCC QA273 (ebook) | LCC QA273 .S6875 2019 (print) | DDC 519.5–dc23
LC record available at https://lccn.loc.gov/2019008498
ISBN 978-1-107-18514-2 Hardback
ISBN 978-1-316-63637-4 Paperback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To my grandchildren Nicholas, Jason, and Evie,
my daughters Stella, Marina, and Alexia, and my
wife Evie for their unconditional love and support
Contents
Preface to the Second Edition page xix
1 An Introduction to Empirical Modeling 1

1.1 Introduction 1
1.2 Stochastic Phenomena: A Preliminary View 3
1.2.1 Chance Regularity Patterns 3
1.2.2 From Chance Regularities to Probabilities 7
1.2.3 Chance Regularity Patterns and Real-World Phenomena 11
1.3 Chance Regularities and Statistical Models 12
1.4 Observed Data and Empirical Modeling 14
1.4.1 Experimental vs. Observational Data 14
1.4.2 Observed Data and the Nature of a Statistical Model 15
1.4.3 Measurement Scales and Data 16
1.4.4 Measurement Scale and Statistical Analysis 18
1.4.5 Cross-Section vs. Time Series, is that the Question? 20
1.4.6 Limitations of Economic Data 22
1.5 Statistical Adequacy 23
1.6 Statistical vs. Substantive Information∗ 25
1.7 Looking Ahead 27
1.8 Questions and Exercises 28
2 Probability Theory as a Modeling Framework 30

2.1 Introduction 30
2.1.1 Primary Objective 30
2.1.2 Descriptive vs. Inferential Statistics 30
2.2 Simple Statistical Model: A Preliminary View 32
2.2.1 The Basic Structure of a Simple Statistical Model 33
2.2.2 The Notion of a Random Variable: A Naive View 34
2.2.3 Density Functions 35
2.2.4 A Random Sample: A Preliminary View 36
2.3 Probability Theory: An Introduction 40
2.3.1 Outlining the Early Milestones of Probability Theory 40
2.3.2 Probability Theory: A Modeling Perspective 42
2.4 A Simple Generic Stochastic Mechanism 42
2.4.1 The Notion of a Random Experiment 42
2.4.2 A Bird’s-Eye View of the Unfolding Story 44
vii
viii Contents
2.5 Formalizing Condition [a]: The Outcomes Set 45

2.5.1 The Concept of a Set in Set Theory 45
2.5.2 The Outcomes Set 45
2.5.3 Special Types of Sets 46
2.6 Formalizing Condition [b]: Events and Probabilities 48
2.6.1 Set-Theoretic Operations 48
2.6.2 Events vs. Outcomes 51
2.6.3 Event Space 51
2.6.4 A Digression: What is a Function? 58
2.6.5 The Mathematical Notion of Probability 59
2.6.6 Probability Space (S,, P(.)) 63
2.6.7 Mathematical Deduction 64
2.7 Conditional Probability and Independence 65
2.7.1 Conditional Probability and its Properties 65
2.7.2 The Concept of Independence Among Events 69
2.8 Formalizing Condition [c]: Sampling Space 70
2.8.1 The Concept of Random Trials 70
2.8.2 The Concept of a Statistical Space 72
2.8.3 The Unfolding Story Ahead 74
3 The Concept of a Probability Model 78

3.1 Introduction 78
3.1.1 The Story So Far and What Comes Next 78
3.2 The Concept of a Random Variable 79
3.2.1 The Case of a Finite Outcomes Set: S = {s1 , s2 , . . . , sn } 80
3.2.2 Key Features of a Random Variable 81
3.2.3 The Case of a Countable Outcomes Set:
S = {s1 , s2 , . . . , sn , . . .} 85
3.3 The General Concept of a Random Variable 86
3.3.1 The Case of an Uncountable Outcomes Set S 86
3.4 Cumulative Distribution and Density Functions 89
3.4.1 The Concept of a Cumulative Distribution Function 89
3.4.2 The Concept of a Density Function 91
3.5 From a Probability Space to a Probability Model 95
3.5.1 Parameters and Moments 97
3.5.2 Functions of a Random Variable 97
3.5.3 Numerical Characteristics of Random Variables 99
3.5.4 Higher Moments 102
3.5.5 The Problem of Moments* 110
3.5.6 Other Numerical Characteristics 112
3.6 Summary 118
Appendix 3.A: Univariate Distributions 121
Contents ix
3.A.1 Discrete Univariate Distributions 121

3.A.2 Continuous Univariate Distributions 123
4 A Simple Statistical Model 130

4.1 Introduction 130
4.1.1 The Story So Far, a Summary 130
4.1.2 From Random Trials to a Random Sample: A First View 130
4.2 Joint Distributions of Random Variables 131
4.2.1 Joint Distributions of Discrete Random Variables 131
4.2.2 Joint Distributions of Continuous Random Variables 133
4.2.3 Joint Moments of Random Variables 136
4.2.4 The n Random Variables Joint Distribution 138
4.3 Marginal Distributions 139
4.4 Conditional Distributions 142
4.4.1 Conditional Probability 142
4.4.2 Conditional Density Functions 143
4.4.3 Continuous/Discrete Random Variables* 146
4.4.4 Conditional Moments 146
4.4.5 A Digression: Other Forms of Conditioning 148
4.4.6 Marginalization vs. Conditioning 150
4.4.7 Conditioning on Events vs. Random Variables 151
4.5 Independence 155
4.5.1 Independence in the Two Random Variable Case 155
4.5.2 Independence in the n Random Variable Case 156
4.6 Identical Distributions and Random Samples 158
4.6.1 Identically Distributed Random Variables 158
4.6.2 A Random Sample of Random Variables 160
4.7 Functions of Random Variables 161
4.7.1 Functions of One Random Variable 161
4.7.2 Functions of Several Random Variables 162
4.7.3 Ordered Sample and its Distributions* 165
4.8 A Simple Statistical Model 166
4.8.1 From a Random Experiment to a Simple Statistical Model 166
4.9 The Statistical Model in Empirical Modeling 167
4.9.1 The Concept of a Statistical Model: A Preliminary View 167
4.9.2 Statistical Identification of Parameters 168
4.9.3 The Unfolding Story Ahead 169
Appendix 4.A: Bivariate Distributions 171
4.A.1 Discrete Bivariate Distributions 171
4.A.2 Continuous Bivariate Distributions 172
5 Chance Regularities and Probabilistic Concepts 176

x Contents
5.1.1 Early Developments in Graphical Techniques 176

5.1.2 Why Do We Care About Graphical Techniques? 177
5.2 The t-Plot and Independence 178
5.3 The t-Plot and Homogeneity 184
5.4 Assessing Distribution Assumptions 189
5.4.1 Data that Exhibit Dependence/Heterogeneity 189
5.4.2 Data that Exhibit Normal IID Chance Regularities 195
5.4.3 Data that Exhibit Non-Normal IID Regularities 196
5.4.4 The Histogram, the Density Function, and Smoothing 201
5.4.5 Smoothed Histograms and Non-Random Samples 206
5.5 The Empirical CDF and Related Graphs* 206
5.5.1 The Concept of the Empirical cdf (ecdf) 207
5.5.2 Probability Plots 208
5.5.3 Empirical Example: Exchange Rate Data 215
5.6 Summary 218
Appendix 5.A: Data – Log-Returns 220
6 Statistical Models and Dependence 222

6.1.1 Extending a Simple Statistical Model 222
6.2 Non-Random Sample: A Preliminary View 224
6.2.1 Sequential Conditioning: Reducing the Dimensionality 225
6.2.2 Keeping an Eye on the Forest! 227
6.3 Dependence and Joint Distributions 228
6.3.1 Dependence Between Two Random Variables 228
6.4 Dependence and Moments 229
6.4.1 Joint Moments and Dependence 229
6.4.2 Conditional Moments and Dependence 232
6.5 Joint Distributions and Modeling Dependence 233
6.5.1 Dependence and the Normal Distribution 234
6.5.2 A Graphical Display: The Scatterplot 236
6.5.3 Dependence and the Elliptically Symmetric Family 240
6.5.4 Dependence and Skewed Distributions 245
6.5.5 Dependence in the Presence of Heterogeneity 257
6.6 Modeling Dependence and Copulas* 258
6.7 Dependence for Categorical Variables 262
6.7.1 Measurement Scales and Dependence 262
6.7.2 Dependence and Ordinal Variables 263
6.7.3 Dependence and Nominal Variables 266
6.8 Conditional Independence 268
6.8.1 The Multivariate Normal Distribution 269
6.8.2 The Multivariate Bernoulli Distribution 271
6.8.3 Dependence in Mixed (Discrete/Continuous) Variables 272
Contents xi
6.9 What Comes Next? 273

7 Regression Models 277

7.2 Conditioning and Regression 279
7.2.1 Reduction and Conditional Moment Functions 279
7.2.2 Regression and Skedastic Functions 281
7.2.3 Selecting an Appropriate Regression Model 288
7.3 Weak Exogeneity and Stochastic Conditioning 292
7.3.1 The Concept of Weak Exogeneity 292
7.3.2 Conditioning on a σ -Field 295
7.3.3 Stochastic Conditional Expectation and its Properties 297
7.4 A Statistical Interpretation of Regression 301
7.4.1 The Statistical Generating Mechanism 301
7.4.2 Statistical vs. Substantive Models, Once Again 304
7.5 Regression Models and Heterogeneity 308
7.6 Summary and Conclusions 310
8 Introduction to Stochastic Processes 315

8.1.1 Random Variables and Orderings 316
8.2 The Concept of a Stochastic Process 318
8.2.1 Defining a Stochastic Process 318
8.2.2 Classifying Stochastic Processes; What a Mess! 320
8.2.3 Characterizing a Stochastic Process 322
8.2.4 Partial Sums and Associated Stochastic Processes 324
8.2.5 Gaussian (Normal) Process: A First View 328
8.3 Dependence Restrictions (Assumptions) 329
8.3.1 Distribution-Based Concepts of Dependence 329
8.3.2 Moment-Based Concepts of Dependence 330
8.4 Heterogeneity Restrictions (Assumptions) 331
8.4.1 Distribution-Based Heterogeneity Assumptions 331
8.4.2 Moment-Based Heterogeneity Assumptions 333
8.5 Building Block Stochastic Processes 335
8.5.1 IID Stochastic Processes 335
8.5.2 White-Noise Process 336
8.6 Markov and Related Stochastic Processes 336
8.6.1 Markov Process 336
8.6.2 Random Walk Processes 338
8.6.3 Martingale Processes 340
8.6.4 Martingale Difference Process 342
8.7 Gaussian Processes 345
xii Contents
8.7.1 AR(p) Process: Probabilistic Reduction Perspective 345

8.7.2 A Wiener Process and a Unit Root [UR(1)] Model 349
8.7.3 Moving Average [MA(q)] Process 352
8.7.4 Autoregressive vs. Moving Average Processes 353
8.7.5 The Brownian Motion Process* 354
8.8 Counting Processes* 360
8.8.1 The Poisson Process 361
8.8.2 Duration (Hazard-Based) Models 363
Appendix 8.A: Asymptotic Dependence and Heterogeneity
Assumptions* 369
8.A.1 Mixing Conditions 369
8.A.2 Ergodicity 370
9 Limit Theorems in Probability 373

9.1.1 Why Do We Care About Limit Theorems? 374
9.1.2 Terminology and Taxonomy 375
9.1.3 Popular Misconceptions About Limit Theorems 376
9.2 Tracing the Roots of Limit Theorems 377
9.2.1 Bernoulli’s Law of Large Numbers: A First View 377
9.2.2 Early Steps Toward the Central Limit Theorem 378
9.2.3 The First SLLN 381
9.2.4 Probabilistic Convergence Modes: A First View 381
9.3 The Weak Law of Large Numbers 383
9.3.1 Bernoulli’s WLLN 383
9.3.2 Poisson’s WLLN 385
9.3.3 Chebyshev’s WLLN 386
9.3.4 Markov’s WLLN 387
9.3.5 Bernstein’s WLLN 388
9.3.6 Khinchin’s WLLN 389
9.4 The Strong Law of Large Numbers 390
9.4.1 Borel’s (1909) SLLN 390
9.4.2 Kolmogorov’s SLLN 391
9.4.3 SLLN for a Martingale 392
9.4.4 SLLN for a Stationary Process 394
9.4.5 The Law of Iterated Logarithm* 395
9.5 The Central Limit Theorem 396
9.5.1 De Moivre–Laplace CLT 397
9.5.2 Lyapunov’s CLT 399
9.5.3 Lindeberg–Feller’s CLT 399
9.5.4 Chebyshev’s CLT 401
9.5.5 Hajek–Sidak CLT 401
Contents xiii
9.5.6 CLT for a Martingale 402

9.5.7 CLT for a Stationary Process 402
9.5.8 The Accuracy of the Normal Approximation 403
9.5.9 Stable and Other Limit Distributions* 404
9.6 Extending the Limit Theorems* 406
9.6.1 A Uniform SLLN* 409
Appendix 9.A: Probabilistic Inequalities 412
9.A.1 Probability 412
9.A.2 Expectation 413
Appendix 9.B: Functional Central Limit Theorem 414
10 From Probability Theory to Statistical Inference 421

10.2 Mathematical Probability: A Brief Summary 422
10.2.1 Kolmogorov’s Axiomatic Approach 422
10.2.2 Random Variables and Statistical Models 422
10.3 Frequentist Interpretation(s) of Probability 423
10.3.1 “Randomness” (Stochasticity) is a Feature of the Real
World 423
10.3.2 Model-Based Frequentist Interpretation of Probability 424
10.3.3 Von Mises’ Frequentist Interpretation of Probability 426
10.3.4 Criticisms Leveled Against the Frequentist Interpretation 427
10.3.5 Kolmogorov Complexity: An Algorithmic Perspective 430
10.3.6 The Propensity Interpretation of Probability 431
10.4 Degree of Belief Interpretation(s) of Probability 432
10.4.1 “Randomness” is in the Mind of the Beholder 432
10.4.2 Degrees of Subjective Belief 432
10.4.3 Degrees of “Objective Belief”: Logical Probability 435
10.4.4 Which Interpretation of Probability? 436
10.5 Frequentist vs. Bayesian Statistical Inference 436
10.5.1 The Frequentist Approach to Statistical Inference 436
10.5.2 The Bayesian Approach to Statistical Inference 440
10.5.3 Cautionary Notes on Misleading Bayesian Claims 443
10.6 An Introduction to Frequentist Inference 444
10.6.1 Fisher and Neglected Aspects of Frequentist Statistics 444
10.6.2 Basic Frequentist Concepts and Distinctions 446
10.6.3 Estimation: Point and Interval 447
10.6.4 Hypothesis Testing: A First View 449
10.6.5 Prediction (Forecasting) 450
10.6.6 Probability vs. Frequencies: The Empirical CDF 450
10.7 Non-Parametric Inference 453
10.7.1 Parametric vs. Non-Parametric Inference 453
xiv Contents
10.7.2 Are Weaker Assumptions Preferable to Stronger Ones? 454

10.7.3 Induction vs. Deduction 457
10.7.4 Revisiting Generic Robustness Claims 458
10.7.5 Inference Based on Asymptotic Bounds 458
10.7.6 Whither Non-Parametric Modeling? 460
10.8 The Basic Bootstrap Method 461
10.8.1 Bootstrapping and Statistical Adequacy 462
11 Estimation I: Properties of Estimators 469

11.2 What is an Estimator? 469
11.3 Sampling Distributions of Estimators 472
11.4 Finite Sample Properties of Estimators 474
11.4.1 Unbiasedness 474
11.4.2 Efficiency: Relative vs. Full Efficiency 475
11.4.3 Sufficiency 480
11.4.4 Minimum MSE Estimators and Admissibility 485
11.5 Asymptotic Properties of Estimators 488
11.5.1 Consistency (Weak) 488
11.5.2 Consistency (Strong) 490
11.5.3 Asymptotic Normality 490
11.5.4 Asymptotic Efficiency 491
11.5.5 Properties of Estimators Beyond the First Two Moments 492
11.6 The Simple Normal Model: Estimation 493
11.7 Confidence Intervals (Interval Estimation) 498
11.7.1 Long-Run “Interpretation” of CIs 499
11.7.2 Constructing a Confidence Interval 499
11.7.3 Optimality of Confidence Intervals 501
11.8 Bayesian Estimation 502
11.8.1 Optimal Bayesian Rules 503
11.8.2 Bayesian Credible Intervals 504
12 Estimation II: Methods of Estimation 510

12.2 The Maximum Likelihood Method 511
12.2.1 The Likelihood Function 511
12.2.2 Maximum Likelihood Estimators 514
12.2.3 The Score Function 517
12.2.4 Two-Parameter Statistical Model 519
12.2.5 Properties of Maximum Likelihood Estimators 524
Contents xv
12.2.6 The Maximum Likelihood Method and its Critics 532

12.3 The Least-Squares Method 534
12.3.1 The Mathematical Principle of Least Squares 534
12.3.2 Least Squares as a Statistical Method 535
12.4 Moment Matching Principle 536
12.4.1 Sample Moments and their Properties 539
12.5 The Method of Moments 543
12.5.1 Karl Pearson’s Method of Moments 543
12.5.2 The Parametric Method of Moments 544
12.5.3 Properties of PMM Estimators 546
Appendix 12.A: Karl Pearson’s Approach 551
13 Hypothesis Testing 553

13.1.1 Difficulties in Mastering Statistical Testing 553
13.2 Statistical Testing Before R. A. Fisher 555
13.2.1 Francis Edgeworth’s Testing 555
13.2.2 Karl Pearson’s Testing 556
13.3 Fisher’s Significance Testing 558
13.3.1 A Closer Look at the p-value 561
13.3.2 R. A. Fisher and Experimental Design 563
13.3.3 Significance Testing: Empirical Examples 565
13.3.4 Summary of Fisher’s Significance Testing 568
13.4 Neyman–Pearson Testing 569
13.4.1 N-P Objective: Improving Fisher’s Significance Testing 569
13.4.2 Modifying Fisher’s Testing Framing: A First View 570
13.4.3 A Historical Excursion 574
13.4.4 The Archetypal N-P Testing Framing 575
13.4.5 Significance Level α vs. the p-value 578
13.4.6 Optimality of a Neyman–Pearson Test 580
13.4.7 Constructing Optimal Tests: The N-P Lemma 586
13.4.8 Extending the Neyman–Pearson Lemma 588
13.4.9 Constructing Optimal Tests: Likelihood Ratio 591
13.4.10 Bayesian Testing Using the Bayes Factor 594
13.5 Error-Statistical Framing of Statistical Testing 596
13.5.1 N-P Testing Driven by Substantively Relevant Values 596
13.5.2 Foundational Issues Pertaining to Statistical Testing 598
13.5.3 Post-Data Severity Evaluation: An Evidential Account 600
13.5.4 Revisiting Issues Bedeviling Frequentist Testing 603
13.5.5 The Replication Crises and Severity 609
13.6 Confidence Intervals and their Optimality 610
13.6.1 Mathematical Duality Between Testing and CIs 610
xvi Contents
13.6.2 Uniformly Most Accurate CIs 612

13.6.3 Confidence Intervals vs. Hypothesis Testing 613
13.6.4 Observed Confidence Intervals and Severity 614
13.6.5 Fallacious Arguments for Using CIs 614
Appendix 13.A: Testing Differences Between Means 620
13.A.1 Testing the Difference Between Two Means 620
13.A.2 What Happens when Var(X1t ) = Var(X2t )? 621
13.A.3 Bivariate Normal Model: Paired Sample Tests 622
13.A.4 Testing the Difference Between Two Proportions 623
13.A.5 One-Way Analysis of Variance 624
14 Linear Regression and Related Models 625

14.1.1 What is a Statistical Model? 625
14.2 Normal, Linear Regression Model 626
14.2.1 Specification 626
14.2.2 Estimation 628
14.2.3 Fitted Values and Residuals 633
14.2.4 Goodness-of-Fit Measures 635
14.2.5 Confidence Intervals and Hypothesis Testing 635
14.2.6 Normality and the LR Model 642
14.2.7 Testing a Substantive Model Against the Data 643
14.3 Linear Regression and Least Squares 648
14.3.1 Mathematical Approximation and Statistical
Curve-Fitting 648
14.3.2 Gauss–Markov Theorem 651
14.3.3 Asymptotic Properties of OLS Estimators 653
14.4 Regression-Like Statistical Models 655
14.4.1 Gauss Linear Model 655
14.4.2 The Logit and Probit Models 655
14.4.3 The Poisson Regression-Like Model 657
14.4.4 Generalized Linear Models 657
14.4.5 The Gamma Regression-Like Model 658
14.5 Multiple Linear Regression Model 658
14.5.1 Estimation 660
14.5.2 Linear Regression: Matrix Formulation 661
14.5.3 Fitted Values and Residuals 662
14.5.4 OLS Estimators and their Sampling Distributions 665
14.6 The LR Model: Numerical Issues and Problems 666
14.6.1 The Problem of Near-Collinearity 666
14.6.2 The Hat Matrix and Influential Observations 673
14.6.3 Individual Observation Influence Measures 674
Contents xvii
14.7 Conclusions 675

Appendix 14.A: Generalized Linear Models 680
14.A.1 Exponential Family of Distributions 680
14.A.2 Common Features of Generalized Linear Models 681
14.A.3 MLE and the Exponential Family 682
Appendix 14.B: Data 683
15 Misspecification (M-S) Testing 685

15.2 Misspecification and Inference: A First View 688
15.2.1 Actual vs. Nominal Error Probabilities 688
15.2.2 Reluctance to Test the Validity of Model Assumptions 691
15.3 Non-Parametric (Omnibus) M-S Tests 694
15.3.1 The Runs M-S Test for the IID Assumptions [2]–[4] 694
15.3.2 Kolmogorov’s M-S Test for Normality ([1]) 695
15.4 Parametric (Directional) M-S Testing 697
15.4.1 A Parametric M-S Test for Independence ([4]) 697
15.4.2 Testing Independence and Mean Constancy ([2] and [4]) 698
15.4.3 Testing Independence and Variance Constancy ([2]
and [4]) 700
15.4.4 The Skewness–Kurtosis Test of Normality 700
15.4.5 Simple Normal Model: A Summary of M-S Testing 701
15.5 Misspecification Testing: A Formalization 703
15.5.1 Placing M-S Testing in a Proper Context 703
15.5.2 Securing the Effectiveness/Reliability of M-S Testing 704
15.5.3 M-S Testing and the Linear Regression Model 705
15.5.4 The Multiple Testing (Comparisons) Issue 706
15.5.5 Testing for t-Invariance of the Parameters 707
15.5.6 Where do Auxiliary Regressions Come From? 707
15.5.7 M-S Testing for Logit/Probit Models 710
15.5.8 Revisiting Yule’s “Nonsense Correlations” 710
15.5.9 Respecification 713
15.6 An Illustration of Empirical Modeling 716
15.6.1 The Traditional Curve-Fitting Perspective 716
15.6.2 Traditional ad hoc M-S Testing and Respecification 718
15.6.3 The Probabilistic Reduction Approach 721
Appendix 15.A: Data 734
References 736
Index 752
Preface to the Second
Edition
The original book, published 20 years ago, has been thoroughly revised with two objectives
in mind. First, to make the discussion more compact and coherent by avoiding repetition and
many digressions. Second, to improve the methodological coherence of the proposed empir-
ical modeling framework by including material pertaining to foundational issues that has
been published by the author over the last 20 years or so in journals on econometrics, statis-
tics, and philosophy of science. In particular, this revised edition brings out more clearly
several crucial distinctions that elucidate empirical modeling, including (a) the statistical
vs. the substantive information/model, (b) the modeling vs. the inference facet of statistical
analysis, (c) testing within and testing outside the boundary of a statistical model, and (d)
pre-data vs. post-data error probabilities. These distinctions shed light on several founda-
tional issues and suggest solutions. In addition, the comprehensiveness of the book has been
improved by adding Chapter 14 on the linear regression and related models.
The current debates on the “replication crises” render the methodological framework
articulated in this book especially relevant for today’s practitioner. A closer look at the
debates (Mayo, 2018) reveals that the non-replicability of empirical evidence problem is,
first and foremost, a problem of untrustworthy evidence routinely published in prestigious
journals. The current focus of that literature on the abuse of significance testing is rather
misplaced, because it is only a part of a much broader problem relating to the mechan-
ical application of statistical methods without a real understanding of their assumptions,
limitations, proper implementation, and interpretation of their results. The abuse and mis-
interpretation of the p-value is just symptomatic of the same uninformed implementation
that contributes majorly to the problem of untrustworthy evidence. Indeed, the same unin-
formed implementation often ensures that untrustworthy evidence is routinely replicated,
when the same mistakes are repeated by equally uninformed practitioners! In contrast to the
current conventional wisdom, it is argued that a major contributor to the untrustworthy evi-
dence problem is statistical misspecification: invalid probabilistic assumptions imposed on
one’s data, another symptom of the same uninformed implementation. The primary objec-
tive of this book is to provide the necessary probabilistic foundation and the overarching
modeling framework for an informed and thoughtful application of statistical methods, as
well as the proper interpretation of their inferential results. The emphasis is placed less on
the mechanics of the application of statistical methods, and more on understanding their
assumptions, limitations, and proper interpretation.
xix
xx Preface to the Second Edition
Key Features of the Book

● It offers a seamless integration of probability theory and statistical inference with a view
to elucidating the interplay between deduction and induction in “learning from data”
about observable phenomena of interest using statistical procedures.
● It develops frequentist modeling and inference from first principles by emphasizing the
notion of a statistical model and its adequacy (the validity of its probabilistic assump-
tions vis-à-vis the particular data) as the cornerstone for reliable inductive inference and
trustworthy evidence.
● It presents frequentist inference as well-grounded procedures whose optimality is
assessed by their capacity to achieve genuine “learning from data.”
● It focuses primarily on the skills and the technical knowledge one needs to be able to
begin with substantive questions of interest, select the relevant data carefully, and pro-
ceed to establish trustworthy evidence for or against hypotheses or claims relating to the
questions of interest. These skills include understanding the statistical information con-
veyed by data plots, selecting appropriate statistical models, as well as validating them
using misspecification testing before any inferences are drawn.
● It articulates reasoned responses to several charges leveled against several aspects of
frequentist inference by addressing the underlying foundational issues, including the
use and abuse of p-values and confidence intervals, Neyman–Pearson vs. Fisher test-
ing, and inference results vs. evidence that have bedeviled frequentist inference since
the 1930s. The book discusses several such foundational issues/problems and proposes
ways to address them using an error statistical perspective grounded in the concept of
severity. Methodological issues discussed in this book include rebuttals to widely used,
ill-thought-out arguments for ignoring statistical misspecification, as well as principled
responses to certain Bayesian criticisms of the frequentist aproach.
● Its methodological perspective differs from the traditional textbook perspective by bring-
ing out the perils of curve-fitting and focusing on the key question: How can empirical
modeling lead to “learning from data” about phenomena of interest by giving rise to
trustworthy evidence?
N O T E: All sections marked with an asterisk (∗) can be skipped at first reading without any
serious interruption in the flow of the discussion.
Acknowledgments
More than any other person, Deborah G. Mayo, my colleague and collaborator on many
foundational issues in statistical inference, has helped to shape my views on several method-
ological issues addressed in this book; for that and the constant encouragement, I’m most
grateful to her. I’m also thankful to Clark Glymour, the other philosopher of science with
whom I had numerous elucidating and creative discussions on many philosophical issues
discoursed in the book. Thanks are also due to Sir David Cox for many discussions that
helped me appreciate the different perspectives on frequentist inference. Special thanks are
also due to my longtime collaborator, Anya McGuirk, who contributed majorly in puzzling
out several thorny issues discussed in this book. I owe a special thanks to Julio Lopez for
Preface to the Second Edition xxi
his insightful comments, as well as his unwavering faith in the coherence and value of the
proposed approach to empirical modeling. I’m also thankful to Jesse Bledsoe for helpful
comments on chapter 13 and Mariusz Kamienski for invaluable help on the front cover
design.
I owe special thanks to several of my former and current students over the last 20 years,
who helped to improve the discussion in this book by commenting on earlier drafts and
finding mistakes and typos. They include Elena Andreou, Andros Kourtellos, Carlos Elias,
Maria Heracleous, Jason Bergtold, Ebere Akobundu, Andreas Koutris, Alfredo Romero,
Niraj Pouydal, Michael Michaelides, Karo Solat, and Mohammad Banasaz.
Symbols
N – set of natural numbers N:={1, 2, ..., n, ...}
R – the set of real numbers; the real line (−∞, ∞)
n times

Rn :=R × R× · · · × R
R+ – the set of positive real numbers; the half real line (0, ∞)
f (x; θ ) – density function of X with parameters θ
F(x; θ ) – cumulative distribution function of X with parameters θ
N(μ, σ 2 ) – Normal distribution with mean μ and variance σ 2
E – Random Experiment (RE)
S – outcomes set (sample space)
– event space (a σ −field)
P(.) – probability set function
σ (X) – minimal sigma-field generated by X
Acronyms
AR(p) – Autoregressive model with p lags
CAN – Consistent, Asymptotically Normal
cdf – cumulative distribution function
CLT – Central Limit Theorem
ecdf – empirical cumulative distributrion function
GM – Generating Mechanism
IID – Indepedent and Identically Distributed
LS – Least-Squares
ML – Maximum Likelihood
M-S – Mis-Specification
N-P – Neyman-Pearson
PMM – Parametric Method of Moments
SLLN – Strong Law Large Numbers
WLLN – Weak Law Large Numbers
UMP – Uniformly Most Powerful
1 An Introduction
to Empirical Modeling
1.1 Introduction
Empirical modeling, broadly speaking, refers to the process, methods, and strategies
grounded on statistical modeling and inference whose primary aim is to give rise to “learn-
ing from data” about stochastic observable phenomena, using statistical models. Real-world
phenomena of interest are said to be “stochastic,” and thus amenable to statistical modeling,
when the data they give rise to exhibit chance regularity patterns, irrespective of whether
they arise from passive observation or active experimentation. In this sense, empirical
modeling has three crucial features:
(a) it is based on observed data that exhibit chance regularities;

(b) its cornerstone is the concept of a statistical model that decribes a probabilistic
generating mechanism that could have given rise to the data in question;
(c) it provides the framework for combining the statistical and substantive informa-
tion with a view to elucidating (understanding, predicting, explaining) phenomena of
interest.
Statistical vs. substantive information. Empirical modeling across different disciplines

involves an intricate blending of substantive subject matter and statistical information. The
substantive information stems from a theory or theories pertaining to the phenomenon of
interest that could range from simple conjectures to intricate substantive (structural) mod-
els. Such information has an important and multifaceted role to play by demarcating the
crucial aspects of the phenomenon of interest (suggesting the relevant variables and data),
as well as enhancing the learning from data when it meliorates the statistical information
without belying it. In contrast, statistical information stems from the chance regularities in
data. Scientific knowledge often begins with substantive conjectures based on subject matter
information, but it becomes knowledge when its veracity is firmly grounded in real-world
data. In this sense, success in “learning from data” stems primarily from a harmonious blend-
ing of these two sources of information into an empirical model that is both statistically and
substantively “adequate”; see Sections 1.5 and 1.6.
1
2 An Introduction to Empirical Modeling
Empirical modeling as curve-fitting. The current traditional perspective on empirical mod-

eling largely ignores the above distinctions by viewing the statistical problem as “quantifying
theoretical relationships presumed true.” From this perspective, empirical modeling is
viewed as a curve-fitting problem, guided primarily by goodness-of-fit. The substantive
model is often imposed on the data in an attempt to quantify its unknown parameters. This
treats the substantive information as established knowledge, and not as tentative conjec-
tures to be tested against data. The end result of curve-fitting is often an estimated model
that is misspecified, both statistically (invalid probabilistic assumptions) and substantively;
it doesn’t elucidate sufficiently the phenomenon of interest. This raises a thorny problem
in philosophy of science known as Duhem’s conundrum (Mayo, 1996), because there is
no principled way to distinguish between the two types of misspecification and apportion
blame. It is argued that the best way to address this impasse is (i) to disentangle the sta-
tistical from the substantive model by unveiling the probabilistic assumptions (implicitly
or explicitly) imposed on the data (the statistical model) and (ii) to separate the modeling
from the inference facet of empirical modeling. The modeling facet includes specifying and
selecting a statistical model, as well as appraising its adequacy (the validity of its proba-
bilistic assumptions) using misspecification testing. The inference facet uses a statistically
adequate model to pose questions of substantive interest to the data. Crudely put, conflating
the modeling with the inference facet is analogous to mistaking the process of constructing a
boat to preset specifications with sailing it in a competitive race; imagine trying to construct
the boat while sailing it in a competitive race.
Early cautionary note. It is likely that some scholars in empirical modeling will mock and
criticize the introduction of new terms and distinctions in this book as “mounds of gratu-
itous jargon,” symptomatic of an ostentatious display of pedantry. As a pre-emptive response
to such critics, allow me to quote R. A. Fisher’s 1931 reply to Arne Fisher’s [American
mathematician/statistician] complaining about his
“introduction in statistical method of some outlandish and barbarous technical terms. They stand
out like quills upon the porcupine, ready to impale the sceptical critic. Where, for instance, did
you get that atrocity, a statistic?”
His serene response was:
I use special words for the best way of expressing special meanings. Thiele and Pearson were
quite content to use the same words for what they were estimating and for their estimates of it.
Hence the chaos in which they left the problem of estimation. Those of us who wish to distinguish
the two ideas prefer to use different words, hence ‘parameter’ and ‘statistic’. No one who does not
feel this need is under any obligation to use them. Also, to Hell with pedantry. (Bennett, 1990,
pp. 311–313) [emphasis added]
A bird’s-eye view of the chapter. The rest of this chapter elaborates on the crucial features
of empirical modeling (a)–(c). In Section 1.2 we discuss the meaning of stochastic observ-
able phenomena and why such phenomena are amenable to empirical modeling. Section 1.3
focuses on the relationship between data from stochastic phenomena and statistical models.
Section 1.4, discusses several important issues relating to observed data, including their dif-
ferent measurement scales, nature, and accuracy. In Section 1.5 we discuss the important
notion of statistical adequacy: whether the postulated statistical model “accounts fully for”
the statistical systematic information in the data. Section 1.6 discusses briefly the connection
between a statistical model and the substantive information of interest.
1.2 Stochastic Phenomena: A Preliminary View

This section provides an intuitive explanation for the notion of a stochastic phenomenon as
it relates to the concept of a statistical model, discussed in the next section.
1.2.1 Chance Regularity Patterns

The chance regularities denote patterns that are usually revealed using a variety of graph-
ical techniques and careful preliminary data analysis. The essence of chance regularity, as
suggested by the term itself, comes in the form of two entwined features:
chance an inherent uncertainty relating to the occurrence of particular outcomes;
regularity discernible regularities associated with an aggregate of many outcomes.
T E R M I N O L O G Y: The term “chance regularity” is used in order to avoid possible confusion
with the more commonly used term “randomness.”
At first sight these two attributes might appear to be contradictory, since “chance” is often
understood as the absence of order and “regularity” denotes the presence of order. However,
there is no contradiction because the “disorder” exists at the level of individual outcomes
and the order at the aggregate level. The two attributes should be viewed as inseparable for
the notion of chance regularity to make sense.
Example 1.1 To get some idea about “chance regularity” patterns, consider the data given
in Table 1.1.
Table 1.1 Observed data
3 10 11 5 6 7 10 8 5 11 2 9 9 6 8 4 7 6 5 12
7 8 5 4 6 11 7 10 5 8 7 5 9 8 10 2 7 3 8 10
11 8 9 5 7 3 4 9 10 4 7 4 6 9 7 6 12 8 11 9
10 3 6 9 7 5 8 6 2 9 6 4 7 8 10 5 8 7 9 6
5 7 7 6 12 9 10 4 8 6 5 4 7 8 6 7 11 7 8 3
A glance at Table 1.1 suggests that the observed data constitute integers between 2 and 12,
but no real patterns are apparent, at least at first sight. To bring out any chance regularity
patterns we use a graph as shown in Figure 1.1, t-plot: {(t, xt ), t = 1, 2, . . . , n}.
The first distinction to be drawn is that between chance regularity patterns and determin-
istic regularities that is easy to detect.
Deterministic regularity. When a t-plot exhibits a clear pattern which would enable one
to predict (guess) the value of the next observation exactly, the data are said to exhibit
deterministic regularity. The easiest way to think about deterministic regularity is to visualize
4 An Introduction to Empirical Modeling
12
10
8
x
1 10 20 30 40 50 60 70 80 90 100
Index
Fig. 1.1 t-Plot of a sequence of 100 observations
1.5
1.0
0.5
0.0
x
–0.5
–1.0
–1.5
1 10 20 30 40 50 60 70 80 90 100
Index
Fig. 1.2 Graph of x = 1.5 cos((π/3)t+(π/3))
the graphs of mathematical functions. If a t-plot of data can be depicted by a mathematical

function, the numbers exhibit deterministic regularity; see Figure 1.2.
In contrast to deterministic regularities, to detect chance patterns one needs to perform a
number of thought experiments.
Thought experiment 1–Distribution regularity. Associate each observation with identical

squares and rotate Figure 1.1 anti-clockwise by 90◦ , letting the squares fall vertically to form
a pile on the x-axis. The pile represents the well-known histogram (see Figure 1.3).
The histogram exhibits a clear triangular shape, reflecting a form of regularity often
associated with stable (unchanging) relative frequencies (RF) expressed as percentages
18
16
Relative frequency (%)

14
12
10
8
6
4
2
0
2 3 4 5 6 7 8 9 10 11 12
x
Fig. 1.3 Histogram of the data in Figure 1.1
(%). Each bar of the histogram represents the frequency of each of the integers 2–12.
For example, since the value 3 occurs five times in this data set, its relative frequency is
RF(3)=5/100 = .05. The relative frequency of the value 7 is RF(7)=17/100 = .17, which is
the highest among the values 2–12. For reasons that will become apparent shortly, we name
this discernible distribution regularity.
[1] Distribution: After a large enough number of trials, the relative frequency of the
outcomes forms a seemingly stable distribution shape.
Thought experiment 2. In Figure 1.1, one would hide the observations beyond a certain
value of the index, say t = 40, and try to guess the next outcome on the basis of the observa-
tions up to t = 40. Repeat this along the x-axis for different index values and if it turns out
that it is more or less impossible to use the previous observations to narrow down the poten-
tial outcomes, conclude that there is no dependence pattern that would enable the modeler
to guess the next observation (within narrow bounds) with any certainty. In this experiment
one needs to exclude the extreme values of 2 and 12, because following these values one
is almost certain to get a value greater and smaller, respectively. This type of predictability
is related to the distribution regularity mentioned above. For reference purposes we name
the chance regularity associated with the unpredictability of the next observation given the
previous observations.
[2] Independence: In a sequence of trials, the outcome of any one trial does not
influence and is not influenced by the outcome of any other.
Thought experiment 3. In Figure 1.1 take a wide enough frame (to cover the spread of the
fluctuations) that is also long enough (roughly less than half the length of the horizontal axis)
and let it slide from left to right along the horizontal axis, looking at the picture inside the
frame as it slides along. In cases where the picture does not change significantly, the data
exhibit the chance regularity we call homogeneity, otherwise heterogeneity is present; see
Another random document with
no related content on Scribd:
by one into the arms of gendarmes below. The palaces along the
Riva were a broad ribbon of color with a binding of black coats and
hats. The wall of San Giorgio fronting the barracks was fringed with
the yellow legs and edged with the white fatigue caps of two
regiments. Even over the roofs and tower of the church itself specks
of sight-seers were spattered here and there, as if the joyous wind in
some mad frolic had caught them up in very glee, and as suddenly
showered them on cornice, sill, and dome.
Beyond all this, away out on the lagoon, toward the islands, the red-
sailed fishing-boats hurried in for the finish, their canvas aflame
against the deepening blue. Over all the sunlight danced and blazed
and shimmered, gilding and bronzing the roof-jewels of San Marco,
flashing from oar blade, brass, and ferro, silvering the pigeons
whirling deliriously in the intoxicating air, making glad and gay and
happy every soul who breathed the breath of this joyous Venetian
day.
None of all this was lost upon the Professor. He stood in the bow
drinking in the scene, sweeping his glass round like a weather-vane,
straining his eyes up the Giudecca to catch the first glimpse of the
coming boats, picking out faces under flaunting parasols, and waving
aloft his yellow rag when some gondola swept by flying Pietro’s
colors, or some boat-load of friends saluted in passing.
Suddenly there came down on the shifting wind, from far up the
Giudecca, a sound like the distant baying of a pack of hounds, and
as suddenly died away. Then the roar of a thousand throats, caught
up by a thousand more about us, broke on the air, as a boatman,
perched on a masthead, waved his hat.
“Here they come! Viva Pietro! Viva Pasquale!—Castellani!—Nicoletti!
—Pietro!”
The dense mass rose and fell in undulations, like a great carpet
being shaken, its colors tossing in the sunlight. Between the thicket
of ferros, away down the silver ribbon, my eye caught two little
specks of yellow capping two white figures. Behind these, almost in
line, were two similar dots of blue; farther away other dots, hardly
distinguishable, on the horizon line.
The gale became a tempest—the roar was deafening; women waved
their shawls in the air; men, swinging their hats, shouted themselves
hoarse. The yellow specks developed into handkerchiefs bound to
the heads of Pietro and his brother Marco; the blues were those of
Pasquale and his mate.
Then, as we strain our eyes, the two tails of the sea-monster twist
and clash together, closing in upon the string of rowers as they
disappear in the dip behind San Giorgio, only to reappear in full
sight, Pietro half a length ahead, straining every sinew, his superb
arms swinging like a flail, his lithe body swaying in splendid,
springing curves, the water rushing from his oar blade, his brother
bending aft in perfect rhythm.
“Pietro! Pietro!” came the cry, shrill and clear, drowning all other
sounds, and a great field of yellow burst into flower all over the
lagoon, from San Giorgio to the Garden. The people went wild. If
before there had been only a tempest, now there was a cyclone. The
waves of blue and yellow surged alternately above the heads of the
throng as Pasquale or Pietro gained or lost a foot. The Professor
grew red and pale by turns, his voice broken to a whisper with
continued cheering, the yellow rag streaming above his head, all the
blood of his ancestors blazing in his face.
The contesting boats surged closer. You could now see the rise and
fall of Pietro’s superb chest, the steel-like grip of his hands, and
could outline the curves of his thighs and back. The ends of the
yellow handkerchief, bound close about his head, were flying in the
wind. His stroke was long and sweeping, his full weight on the oar;
Pasquale’s stroke was short and quick, like the thrust of a spur.
Now they are abreast. Pietro’s eyes are blazing—Pasquale’s teeth
are set. Both crews are doing their utmost. The yells are demoniac.
Even the women are beside themselves with excitement.
Suddenly, when within five hundred yards of the goal, Pasquale
turns his head to his mate; there is an answering cry, and then, as if
some unseen power had lent its strength, Pasquale’s boat shoots
half a length ahead, slackens, falls back, gains again, now an inch,
now a foot, now clear of Pietro’s bow, and on, on, lashing the water,
surging forward, springing with every gain, cheered by a thousand
throats, past the red tower of San Giorgio, past the channel of spiles
off the Garden, past the red buoy near the great warship,—one
quick, sustained, blistering stroke,—until the judge’s flag drops from
his hand, and the great race is won.
“A true knight, a gentleman every inch of him,” called out the
Professor, forgetting that he had staked all his soldi on Pietro. “Fairly
won, Pasquale.”
In the whirl of the victory, I had forgotten Pietro, my gondolier of the
morning. The poor fellow was sitting in the bow of his boat, his head
in his hands, wiping his forehead and throat, the tears streaming
down his cheeks. His brother sat beside him. In the gladness and
disappointment of the hour, no one of the crowd around him seemed
to think of the hero of five minutes before. Not so Giorgio, who was
beside himself with grief over Pietro’s defeat, and who had not taken
his eyes from his face. In an instant more he sprang forward, calling
out, “No! no! Brava Pietro!” Espero joining in as if with a common
impulse, and both forcing their gondolas close to Pietro’s.
A moment more and Giorgio was over the rail of Pietro’s boat,
patting his back, stroking his head, comforting him as you would
think only a woman could—but then you do not know Giorgio. Pietro
lifted up his face and looked into Giorgio’s eyes with an expression
so woe-begone, and full of such intense suffering, that Giorgio
instinctively flung his arm around the great, splendid fellow’s neck.
Then came a few broken words, a tender caressing stroke of
Giorgio’s hand, a drawing of Pietro’s head down on his breast as if it
had been a girl’s, and then, still comforting him—telling him over and
over again how superbly he had rowed, how the next time he would
win, how he had made a grand second—
Giorgio bent his head—and kissed him.
When Pietro, a moment later, pulled himself together and stood erect
in his boat, with eyes still wet, the look on his face was as firm and
determined as ever.
Nobody laughed. It did not shock the crowd; nobody thought Giorgio
unmanly or foolish, or Pietro silly or effeminate. The infernal Anglo-
Saxon custom of always wearing a mask of reserve, if your heart
breaks, has never reached these people.
As for the Professor, who looked on quietly, I think—yes, I am quite
sure—that a little jewel of a tear squeezed itself up through his
punctilious, precise, ever exact and courteous body, and glistened
long enough on his eyelids to wet their lashes. Then the bright sun
and the joyous wind caught it away. Dear old relic of a by-gone time!
How gentle a heart beats under your well-brushed, threadbare coat!
SOME VENETIAN CAFFÈS
VERY one in Venice has his own particular caffè, according
to his own particular needs, sympathies, or tastes. All the
artists, architects, and musicians meet at Florian’s; all the
Venetians go to the Quadri; the Germans and late
Austrians, to the Bauer-Grünwald; the stay-over-nights, to the
Oriental on the Riva; the stevedores, to the Veneta Marina below the
Arsenal; and my dear friend Luigi and his fellow-tramps, to a little
hole in the wall on the Via Garibaldi.
A LITTLE HOLE IN THE WALL ON THE VIA GARIBALDI

These caffès are scattered everywhere, from the Public Garden to
the Mestre bridge; all kinds of caffès for all kinds of people—rich, not
so rich, poor, poorer, and the very poorest. Many of them serve only
a cup of coffee, two little flat lumps of sugar, a hard, brown roll, and a
glass of water—always a glass of water. Some add a few syrups and
cordials, with a siphon of seltzer. Others indulge in the cheaper
wines of the country, Brindisi, Chianti, and the like, and are then
known as wine-shops. Very few serve any spirits, except a spoonful
of cognac with the coffee. Water is the universal beverage, and in
summer this is cooled by ice and enriched by simple syrups of
peach, orange, and raspberry. Spirits are rarely taken and
intemperance is practically unknown. In an experience of many
years, I have not seen ten drunken men,—never one drunken
woman,—and then only in September, when the strong wine from
Brindisi is brought in bulk and sold over the boat’s rail, literally by the
bucket, to whoever will buy.
In the ristoranti—caffès, in our sense—is served an array of eatables
that would puzzle the most expert of gourmands. There will be
macaroni, of course, in all forms, and risotto in a dozen different
ways, and soups with weird, uncanny little devil-fish floating about in
them, and salads of every conceivable green thing that can be
chopped up in a bowl and drowned in olive oil; besides an
assortment of cheeses with individualities of perfume that beggar
any similar collection outside of Holland.
Some of these caffès are so much a part of Venice and Venetian life,
that you are led to believe that they were founded by the early Doges
and are coëval with the Campanile or the Library. Somebody, of
course, must know when they first began setting out tables on the
piazza in front of Florian’s, or at the Quadri opposite, or yet again at
the Caffè al Cavallo, near San Giovanni e Paulo, and at scores of
others; but I confess I do not. If you ask the head waiter, who really
ought to know (for he must have been born in one of the upper
rooms—he certainly never leaves the lower ones), he shrugs his
shoulders in a hopeless way and sheds the inquiry with a despairing
gesture, quite as if you had asked who laid out San Marco, or who
drove the piles under Saint Theodore.
There is, I am convinced, no real, permanent, steady proprietor in
any of these caffès—none that one ever sees. There must be, of
course, somebody who assumes ownership, and who for a time
really believes that he has a proprietary interest in the chairs and
tables about him. After a while, however, he gets old and dies, and is
buried over in Campo Santo, and even his name is forgotten. When
this happens, and it is eminently proper that it should, another tenant
takes possession, quite as the pigeons do of an empty carving over
the door of the king’s palace.
But the caffè keeps on: the same old marble-top tables; the same old
glass-covered pictures, with the impossible Turkish houris listening
to the improbable gentleman in baggy trousers; the same serving-
counter, with the row of cordials in glass bottles with silver stoppers.
The same waiters, too, hurry about—they live on for centuries—
wearing the same coats and neckties, and carrying the same
napkins. I myself have never seen a dead waiter, and, now I happen
to think of it, I have never heard of one.
The head waiter is, of course, supreme. He it is who adds up on his
fingers the sum of your extravagances, who takes your money and
dives down into his own pocket for the change. He and his assistants
are constantly running in and out, vanishing down subterranean
stairs, or disappearing through swinging doors, with the agility of
Harlequin; you never know where or why, until they pop out again,
whirling trays held high over their heads, or bearing in both hands
huge waiters loaded with dishes.
The habitués of these caffès are as interesting as the caffès
themselves. The Professor comes, of course; you always know
where to find him. And the youthful Contessa! She of the uncertain
age, with hair bleached to a light law-calf, and a rose-colored veil!
And here comes, too, every distinguished or notorious person of high
or low degree at the moment in Venice; you have only to take a chair
at Florian’s and be patient—they are sure to appear before the music
is over. There is the sister of the Archduke, with the straight-backed,
pipe-stem-legged officer acting as gentleman-in-waiting; and he
does wait, standing bolt upright like a sergeant on dress parade,
sometimes an hour, for her to sit down. There is the Spanish
Grandee, with a palace for the season (an upper floor with an
entrance on a side canal), whose gondoliers wear flaming scarlet,
with a crest embossed on brass dinner-plates for arm ornaments;
one of these liveried attendants always dogs the Grandee to the
caffè, so as to be ready to pull his chair out when his Excellency sits
down. Then there are the Royal Academician, in gray tweed
knickerbockers, traveling incognito with two friends; the fragments of
an American linen-duster brigade, with red guide-books and faces, in
charge of a special agent; besides scores of others of every
nationality and rank. They are all at Florian’s some time during the
day.
You will see there, too, if you are familiar with the inside workings of
a favorite caffè, an underground life of intrigue or mystery, in which
Gustavo or Florio has a hand—often upon a billet-doux concealed
within the folds of a napkin; not to mention the harmless distribution,
once in a while, of smuggled cigarettes fresh from Cairo.
Poor Gustavo! The government brought him to book not long ago.
For many years he had supplied his patrons, and with delicious
Egyptians, too! One night Gustavo disappeared, escorted by two
gendarmes from the Department of Justice. Next morning the judge
said: “Whereas, according to the accounts kept by the Department of
Customs, the duties and expenses due the king on the cigarettes
unlawfully sold by the prisoner for years past aggregate two
thousand three hundred and ten lire; and whereas, the savings of the
prisoner for ten years past, and at the moment deposited to his
individual credit in the Banco Napoli, amount to exactly two thousand
three hundred and ten lire; therefore, it is ordered, that a sight draft
for the exact amount be drawn in favor of the king.” This would
entitle Gustavo to the pure air of the piazza; otherwise?—well,
otherwise not. Within a week Gustavo was again whirling his tray—a
little grayer, perhaps, and a little wiser; certainly poorer. Thus does a
tyrannical government oppress its people!
These caffès of the piazza, with their iced carafes, white napkins,
and little silver coffee-pots, are the caffès of the rich.
The caffè of the poor is sometimes afloat. No matter how early you
are out in the morning, this floating caffè—the cook-boat—has its fire
lighted, and the savory smell of its cuisine drifts over the lagoon, long
before your gondola rounds the Dogana. When you come alongside
you find a charcoal brazier heating a pan of savory fish and a large
pot of coffee, and near by a basketful of rolls, fresh and warm, from a
still earlier baker. There are peaches, too, and a hamper of figs. The
cook-boat is tended by two men; one cooks and serves, and the
other rows, standing in the stern, looking anxiously for customers,
and calling out in stentorian tones that all the delicacies of the
season are now being fried, broiled, and toasted, and that for the
infinitesimal sum of ten soldi you can breakfast like a doge.
If you are just out of the lagoon, your blood tingling with the touch of
the sea, your face aglow with your early morning bath, answer the
cry of one of these floating kitchens, and eat a breakfast with the
rising sun lighting your forehead and the cool breath of the lagoon
across your cheek. It may be the salt air and the early plunge that
make the coffee so savory, the fish and rolls so delicious, and the
fruit so refreshing; or it may be because the fish were wriggling in the
bottom of the boat half an hour earlier, the coffee only at the first
boiling, and the fruit, bought from a passing boat, still damp with the
night’s dew!
The caffè of the poorest is wherever there is a crowd. It generally
stands on three iron legs under one of the trees down the Via
Garibaldi, or over by the landing of the Dogana, or beneath the
shade of some awning, or up a back court. The old fellow who bends
over the hot earthen dish, supported on these legs, slowly stirring a
mess of kidneys or an indescribable stew, is cook, head waiter, and
proprietor all in one. Every now and then he fishes out some delicate
tidbit—a miniature octopus, perhaps (called fulpe), a little sea-horror,
all legs and claws, which he sprawls out on a bit of brown paper and
lays on the palm of your left hand, assuming, clearly, that you have
all the knives and forks that you need, on your right.
Once in a while a good Bohemian discovers some out-of-the-way
place up a canal or through a twisted calle that delights him with its
cuisine, its cellar, or its cosiness, and forever after he preëmpts it as
his caffè. I know half a dozen such discoveries—one somewhere
near San Giorgio degli Schiavoni, where the men play bowls in a
long, narrow alley, under wide-spreading trees, cramped up between
high buildings; and another, off the Merceria, where the officers
smoke and lounge; and still another, quite my own—the Caffè
Calcina. This last is on the Rio San Vio, and looks out on the
Giudecca, just below San Rosario. You would never suspect it of
being a caffè at all, until you had dodged under the little roof of the
porch to escape the heat, and opening the side door found yourself
in a small, plainly furnished room with little marble-top tables, each
decorated with a Siamese-twin salt-cellar holding a pinch of salt and
of pepper. Even then it is a very common sort of caffè, and not at all
the place you would care to breakfast in twice; that is, not until you
had followed the demure waiter through a narrow passage and out
into a square patio splashed with yellow-green light and cooled by
overlacing vines. Then you realize that this same square patch of
ground is one of the few restful spots of the wide earth.
It is all open to the sky except for a great arbor of grape leaves
covering the whole area, beneath which, on the cool, moist ground,
stand half a dozen little tables covered with snow-white cloths. At
one side is a shelter, from behind which come certain mysterious
noises of fries and broils. All about are big, green-painted boxes of
japonicas, while at one end the oleanders thrust their top branches
through the overhanging leaves of the arbor, waving their blossoms
defiantly in the blazing sun. Beneath this grateful shelter you sit and
loaf and invite your soul, and your best friend, too, if he happens to
be that sort of a man.
After having congratulated yourself on your discovery and having
become a daily habitué of the delightful patio, you find that you have
really discovered the Grand Canal or the Rialto bridge. To your great
surprise, the Caffè Calcina has been the favorite resort of good
Bohemians for nearly a century. You learn that Turner painted his
sunset sketches from its upper windows, and that dozens of more
modern English painters have lived in the rooms above; that Whistler
and Rico and scores of others have broken bread and had
toothsome omelets under its vines; and, more precious than all, that
Ruskin and Browning have shared many a bottle of honest Chianti
with these same oleanders above their heads, and this, too, in the
years when the Sage of Brantwood was teaching the world to love
his Venice, and the great poet was singing songs that will last as
long as the language.
ON THE HOTEL STEPS
F you drink your early coffee as I do, in the garden under the
oleanders, overlooking the water-landing of the hotel, and
linger long enough over your fruit, you will conclude before
many days that a large part of the life of Venice can be
seen from the hotel steps. You may behold the great row of gondolas
at the traghetto near by, ranged side by side, awaiting their turn, and
here and there, tied to the spiles outside the line, the more fortunate
boats whose owners serve some sight-seer by the week, or some
native padrone by the month, and are thus free of the daily routine of
the traghetto, and free, too, from our old friend Joseph’s summoning
voice.
You will be delighted at the good-humor and good-fellowship which
animate this group of gondoliers; their ringing songs and hearty
laughter; their constant care of the boats, their daily sponging and
polishing; and now and then, I regret to say, your ears will be
assailed by a quarrel, so fierce, so loud, and so full of vindictive
energy, that you will start from your seat in instant expectation of the
gleam of a stiletto, until by long experience you learn how harmless
are both the bark and bite of a gondolier, and how necessary as a
safety-valve, to accused and accuser as well, is the unlimited air-
space of the Grand Canal.
You will also come into closer contact with Joseph, prince among
porters, and patron saint of this Traghetto of Santa Salute. There is
another Saint, of course, shaded by its trellised vines, framed in
tawdry gilt, protected from the weather by a wooden hood, and
lighted at night by a dim lamp hanging before it—but, for all that,
Joseph is supreme as protector, refuge, and friend.
Joseph, indeed, is more than this. He is the patron saint and father
confessor of every wayfarer, of whatever tongue. Should a copper-
colored gentleman mount the steps of the hotel landing, attired in
calico trousers, a short jacket of pea-green silk, and six yards of bath
toweling about his head, Joseph instantly addresses him in broken
Hindostanee, sending his rattan chairs and paper boxes to a room
overlooking the shady court, and placing a boy on the rug outside,
ready to spring when the copper-colored gentleman claps his hands.
Does another distinguished foreigner descend from the gondola,
attended by two valets with a block-tin trunk, half a score of hat-
boxes, bags, and bundles, four umbrellas, and a dozen sticks,
Joseph at once accosts him in most excellent English, and has
ordered a green-painted tub rolled into his room before he has had
time to reach the door of his apartment. If another equally
distinguished traveler steps on the marble slab, wearing a Bond
Street ulster, a slouch hat, and a ready-made summer suit, with
yellow shoes, and carrying an Alpine staff (so useful in Venice)
branded with illegible letters chasing each other spirally up and down
the wooden handle, Joseph takes his measure at a glance. He
knows it is his first trip “en Cook,” and that he will want the earth, and
instantly decides that so far as concerns himself he shall have it,
including a small, round, convenient little portable which he
immediately places behind the door to save the marble hearth. So
with the titled Frenchman, wife, maid, and canary bird; the haughty
Austrian, his sword in a buckskin bag; the stolid German with the
stout helpmate and one satchel, or the Spaniard with two friends and
no baggage at all.
Joseph knows them all—their conditions, wants, economics,
meannesses, escapades, and subterfuges. Does he not remember
how you haggled over the price of your room, and the row you made
when your shoes were mixed up with the old gentleman’s on the
floor above? Does he not open the door in the small hours, when
you slink in, the bell sounding like a tocsin at your touch? Is he not
rubbing his eyes and carrying the candle that lights you down to the
corridor door, the only exit from the hotel after midnight, when you
had hoped to escape by the garden, and dare not look up at the
balcony above?
Here also you will often meet the Professor. Indeed, he is
breakfasting with me in this same garden this very morning. It is the
first time I have seen him since the memorable day of the regatta,
when Pasquale won the prize and the old fellow lost his soldi.
He has laid aside his outing costume—the short jacket, beribboned
hat, and huge field-glass—and is gracing my table clothed in what he
is pleased to call his “garb of tuition,” worn to-day because of a pupil
who expects him at nine o’clock; “a horrid old German woman from
Prague,” he calls her. This garb is the same old frock-coat of many
summers, the well-ironed silk hat, and the limp glove dangling from
his hand or laid like a crumpled leaf on the cloth beside him. The
coat, held snug to the waist by a single button, always bulges out
over the chest, the two frogs serving as pockets. From these depths,
near the waist-line, the Professor now and then drags up a great silk
handkerchief, either red or black as the week’s wash may permit, for
I have never known of his owning more than two!
To-day, below the bulge of this too large handkerchief swells yet
another enlargement, to which my guest, tapping it significantly with
his finger-tips, refers in a most mysterious way as “a very great
secret,” but without unbosoming to me either its cause or its mystery.
When the cigarettes are lighted he drops his hand deep into his one-
buttoned coat, unloads the handkerchief, and takes out a little
volume bound in vellum, a book he had promised me for weeks. This
solves the mystery and effaces the bulge.
One of the delights of knowing the Professor well is to see him
handle a book that he loves. He has a peculiar way of smoothing the
sides before opening it, as one would a child’s hand, and of always
turning the leaves as though he were afraid of hurting the back,
caressing them one by one with his fingers, quite as a bird plumes its
feathers. And he is always bringing a new book to light; one of his
charming idiosyncrasies is the hunting about in odd corners for just
such odd volumes.
“Out of print now, my dear fellow. You can’t buy it for money. This is
the only copy in Venice that I could borrow for love. See the chapters
on these very fellows—these gondoliers,” pointing to the traghetto.
“Sometimes, when I hear their quarrels, I wonder if they ever
remember that their guild is as old as the days of the Doges, a fossil
survival, unique, perhaps, in the history of this or of any other
country.”
While the Professor nibbles at the crescents and sips his coffee,
pausing now and then to read me passages taken at random from
the little volume in his hands, I watch the procession of gondolas
from the traghetto, like a row of cabs taking their turn, as Joseph’s “a
una” or “due” rings out over the water. One after another they steal
noiselessly up and touch the water-steps, Joseph helping each party
into its boat: the German Baroness with the two poodles and a silk
parasol; the poor fellow from the Engadine, with the rugs and an
extra overcoat, his mother’s arm about him—not many more
sunshiny days for him; the bevy of joyous young girls in summer
dresses and sailor hats, and the two college boys in white flannels,
the chaperone in the next boat. “Ah, these sweet young Americans,
these naïve countrywomen of yours!” whispers the Professor; “how
exquisitely bold!” Last, the painter, with his trap and a big canvas,
which he lifts in as carefully as if it had a broken rib, and then turns
quickly face in; “an old dodge,” you say to yourself; “unfinished, of
course!”
Presently a tall, finely formed gondolier in dark blue, with a red sash,
whirls the ferro of his boat close to the landing-steps, and a graceful,
dignified woman, past middle life, but still showing traces of great
beauty, steps in, and sinks upon the soft cushions.
The Professor rises like a grand duke receiving a princess, brings
one arm to a salute, places the other over his heart, and makes a
bow that carries the conviction of profound respect and loyalty in its
every curve. The lady acknowledges it with a gracious bend of her
head, and a smile which shows her appreciation of its sincerity.
“An English lady of rank who spends her Octobers here,” says the
Professor, when he regains his seat. He had remained standing until
the gondola had disappeared—such old-time observances are part
of his religion.
“Did you notice her gondolier? That is Giovanni, the famous
oarsman. Let me tell you the most delicious story! Oh, the childish
simplicity of these men! You would say, would you not, that he was
about forty years of age? You saw, too, how broad and big he was?
Well, mon ami, not only is he the strongest oarsman in Venice, but
he has proved it, for he has won the annual regatta, the great one on
the Grand Canal, for five consecutive summers! This, you know,
gives him the title of ‘Emperor.’ Now, there is a most charming
Signora whom he has served for years,—she always spends her
summers here,—whom, I assure you, Giovanni idolizes, and over
whom he watches exactly as if she were both his child and his
queen. Well, one day last year,” here the Professor’s face cracked
into lines of suppressed mirth, “Giovanni asked for a day’s leave,
and went over to Mestre to bid good-by to some friends en route for
Milan. The Brindisi wine—the vina forte—oh, that devilish wine! you
know it!—had just reached Mestre. It only comes in September, and
lasts but a few weeks. Of course Giovanni must have his grand
outing, and three days later Signor Giovanni-the-Strong presented
himself again at the door of the apartment of his Signora, sober, but
limp as a rag. The Signora, grand dame as she was, refused to see
him, sending word by her maid that she would not hear a word from
him until the next day. Now, what do you think this great strong fellow
did? He went home, threw himself on the bed, turned his face to the
wall, and for half the night cried like a baby! Think of it! like a baby!
His wife could not get him to eat a mouthful.
“The next day, of course, the Signora forgave him. There was
nothing else to be done, for, as she said to me afterwards, ‘What?
Venice without Giovanni! Mon Dieu!’”
The Professor throws away the end of his last cigarette and begins
gathering up his hat and the one unmated, lonely glove. No living
soul ever yet saw him put this on. Sometimes he thrusts in his two
fingers, as if fully intending to bury his entire hand, and then you see
an expression of doubt and hesitancy cross his face, denoting a
change of mind, as he crumples it carelessly, or pushes it into his
coat-tail pocket to keep company with its fictitious mate.
At this moment Espero raises his head out of his gondola
immediately beneath us. Everything is ready, he says: the sketch
trap, extra canvas, fresh siphon of seltzer, ice, fiasco of Chianti,
Gorgonzola, all but the rolls, which he will get at the baker’s on our
way over to the Giudecca, where I am to work on the sketch begun
yesterday.
“Ah, that horrid old German woman from Prague!” sighs the
Professor. “If I could only go with you!”
OPEN-AIR MARKETS
OMETIMES, in early autumn, on the lagoon behind the
Redentore, you may overtake a curious craft, half barge,
half gondola, rowed by a stooping figure in cowl and frock.
Against the glow of the fading twilight this quaint figure, standing in
the stern of his flower-laden boat, swaying to the rhythm of his oar,
will recall so vividly the time when that other
“Dumb old servitor ... went upward with the flood,”
that you cannot help straining your eyes in a vain search for the fair
face of the lily maid of Astolat hidden among the blossoms. Upon
looking closer you discover that it is only the gardener of the convent
grounds, on his way to the market above the Rialto.
PONTE PAGLIA ... NEXT THE BRIDGE OF SIGHS

If you continue on, crossing the Giudecca, or if you happen to be
coming from Murano or the Lido, you will pass dozens of other
boats, loaded to the water’s edge with baskets upon baskets of
peaches, melons, and figs, or great heaps of green vegetables,
dashed here and there with piles of blood-red tomatoes. All these
boats are pointing their bows towards the Ponte Paglia, the bridge
on the Riva between the Doges’ Palace and the prison, the one next
the Bridge of Sighs. Here, in the afternoons preceding market days,
they unship their masts or rearrange their cargoes, taking off the top
baskets if too high to clear the arch. Ponte Paglia is the best point of
entrance from the Grand Canal, because it is the beginning of that
short cut, through a series of smaller canals, to the fruit market
above the Rialto bridge. The market opens at daybreak.
Many of these boats come from Malamocco, on the south, a small
island this side of Chioggia, and from beyond the island known as
the Madonna of the Seaweed, named after a curious figure sheltered
by a copper umbrella. Many of them come from Torcello, that most
ancient of the Venetian settlements, and from the fruit-raising country
back of it, for all Torcello is one great orchard, with every landing-
wharf piled full of its products. Here you can taste a fig so delicately
ripe that it fairly melts in your mouth, and so sensitive that it withers
and turns black almost with the handling. Here are rose-pink
peaches the size of small melons, and golden melons the size of
peaches. Here are pomegranates that burst open from very
lusciousness, and white grapes that hang in masses, and melons
and plums in heaps, and all sorts of queer little round things that you
never taste but once, and never want to taste again.
These fruit gardens and orchards in the suburbs of Venice express
the very waste and wantonness of the climate. There is no order in
setting out the fruit, no plan in growing, no system in gathering. The
trees thrive wherever they happen to have taken root—here a peach,
here a pear, there a pomegranate. The vines climb the trunks and
limbs, or swing off to tottering poles and crumbling walls. The
watermelons lie flat on their backs in the blazing sun, flaunting their
big leaves in your face, their tangled creepers in everybody’s way
and under everybody’s feet. The peaches cling in matted clusters,
and the figs and plums weigh down the drooping branches.
If you happen to have a lira about you, and own besides a bushel
basket, you can exchange the coin for that measure of peaches. Two
lire will load your gondola half full of melons; three lire will pack it
with grapes; four lire—well, you must get a larger boat.
When the boats are loaded at the orchards and poled through the
grass-lined canals, reaching the open water of the lagoon, escaping
the swarms of naked boys begging backsheesh of fruit from their
cargoes, you will notice that each craft stops at a square box,
covered by an awning and decorated with a flag, anchored out in the
channel, or moored to a cluster of spiles. This is the Dogana of the
lagoon, and every basket, crate, and box must be inspected and
counted by the official in the flat cap with the tarnished gilt band, who
commands this box of a boat, for each individual peach, plum, and
pear must help pay its share of the public debt.
This floating custom-house is one of many beads, strung at intervals
a mile apart, completely encircling Venice. It is safe to say that
nothing that crows, bleats, or clucks, nothing that feeds, clothes, or is
eaten, ever breaks through this charmed circle without leaving some
portion of its value behind. This creditor takes its pound of flesh the
moment it is due, and has never been known to wait.
Where the deep-water channels are shifting, and there is a
possibility of some more knowing and perhaps less honest market
craft slipping past in the night, a government deputy silently steals
over the shallow lagoon in a rowboat, sleeping in his blanket, his
hand on his musket, and rousing at the faintest sound of rowlock or
sail. Almost hourly one of these night-hawks overhauls other strollers
of the lagoon in the by-passages outside the city limits—some
smuggler, with cargo carefully covered, or perhaps a pair of lovers in
a gondola with too closely drawn tenda. There is no warning sound
to the unwary; only the gurgle of a slowly-moving oar, then the
muzzle of a breech-loader thrust in one’s eyes, behind which frowns
an ugly, determined face, peering from out the folds of a heavy boat-
cloak. It is the deputy’s way of asking for smuggled cigarettes, but it
is so convincing a way as to admit of no discussion. Ever afterward
the unfortunate victim, if he be of honest intent, cannot only detect a
police-boat from a fishing yawl, but remembers also to keep a light
burning in his lamp-socket forward, as evidence of his honesty.
When the cargoes of the market boats are inspected, the duties
paid, and the passage made under Ponte Paglia, or through the
many nameless canals if the approach is made from the Campo
Santo side of the city, the boats swarm up to the fruit market above
the Rialto, rounding up one after another, and discharging their
contents like trucks at a station, the men piling the baskets in great
mounds on the broad stone quay.
THE FRUIT MARKET ABOVE THE RIALTO

After the inhabitants have pounced upon these heaps and mounds
and pyramids of baskets and crates, and have carried them away,
the market is swept and scoured as clean as a china plate, not even
a peach-pit being left to tell the tale of the morning. Then this greater
market shrinks into the smaller one, the little fruit market of the
Rialto, which is never closed, day or night.
This little market, or, rather, the broad street forming its area,—broad
for this part of Venice,—is always piled high with the products of

PDF Probability Theory and Statistical Inference Empirical Modeling With Observational Data 2Nd Edition Spanos A Ebook Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PDF Probability Theory and Statistical Inference Empirical Modeling With Observational Data 2Nd Edition Spanos A Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

Probability theory and statistical

inference empirical modeling with

Probability Theory And Statistical Inference Empirical

Probably Not: Future Prediction Using Probability And

Exact Statistical Inference for Categorical Data 1st

Applied statistical inference with MINITAB Second

Theory of stochastic objects: probability, stochastic

Statistical independence in probability analysis number

Computer Age Statistical Inference Algorithms Evidence

Computer Age Statistical Inference Algorithms Evidence

Doubt over the trustworthiness of published empirical results is not unwarranted

ARIS SPANOS is Wilson E. Schmidt Professor of Economics at Virginia Poly-

Cambridge University Press is part of the University of Cambridge.

Preface to the Second Edition page xix

1 An Introduction to Empirical Modeling 1

2 Probability Theory as a Modeling Framework 30

2.5 Formalizing Condition [a]: The Outcomes Set 45

3 The Concept of a Probability Model 78

3.A.1 Discrete Univariate Distributions 121

4 A Simple Statistical Model 130

5 Chance Regularities and Probabilistic Concepts 176

5.1.1 Early Developments in Graphical Techniques 176

6 Statistical Models and Dependence 222

6.9 What Comes Next? 273

7 Regression Models 277

8 Introduction to Stochastic Processes 315

8.7.1 AR(p) Process: Probabilistic Reduction Perspective 345

9 Limit Theorems in Probability 373

9.5.6 CLT for a Martingale 402

10 From Probability Theory to Statistical Inference 421

10.7.2 Are Weaker Assumptions Preferable to Stronger Ones? 454

11 Estimation I: Properties of Estimators 469

12 Estimation II: Methods of Estimation 510

12.2.6 The Maximum Likelihood Method and its Critics 532

13 Hypothesis Testing 553

13.6.2 Uniformly Most Accurate CIs 612

14 Linear Regression and Related Models 625

14.7 Conclusions 675

15 Misspecification (M-S) Testing 685

Key Features of the Book

(a) it is based on observed data that exhibit chance regularities;

Statistical vs. substantive information. Empirical modeling across different disciplines

Empirical modeling as curve-fitting. The current traditional perspective on empirical mod-

1.2 Stochastic Phenomena: A Preliminary View

1.2.1 Chance Regularity Patterns

Table 1.1 Observed data

Fig. 1.1 t-Plot of a sequence of 100 observations

Fig. 1.2 Graph of x = 1.5 cos((π/3)t+(π/3))

the graphs of mathematical functions. If a t-plot of data can be depicted by a mathematical

Thought experiment 1–Distribution regularity. Associate each observation with identical

Relative frequency (%)

Fig. 1.3 Histogram of the data in Figure 1.1

A LITTLE HOLE IN THE WALL ON THE VIA GARIBALDI

PONTE PAGLIA ... NEXT THE BRIDGE OF SIGHS

THE FRUIT MARKET ABOVE THE RIALTO

You might also like