Making Social Sciences More Scientific: The Need for Predictive Models

Cover......Page 1
Title Page......Page 5
Foreword: Statistical Versus Scientific Inferences......Page 7
Preface......Page 9
Contents......Page 13
List of Figures......Page 15
List of Tables......Page 17
Part I: The Limitations of Descriptive Methodology......Page 19
1 Why Social Sciences Are Not Scientific Enough......Page 21
Omitting One-Half of the Scientific Method......Page 23
Description and Prediction......Page 25
Reversing the Roles of Scientist and Statistician......Page 27
We Can Do Better than That......Page 28
Previous Attempts to Make Social Science More of a Science......Page 30
2 Can Social Science Approaches Find the Law of Gravitation?......Page 32
James McGregor’s Question......Page 33
The Universal Law of Gravitation......Page 34
The Test......Page 35
The Negative Outcome......Page 37
Does It Matter?......Page 38
Typical Analyses of Gravitation-Like Data Offered by Social Scientists......Page 39
3 How to Construct Predictive Models: Simplicity and Nonabsurdity......Page 41
Galileo and the Peasant of Tuscany......Page 42
Directional Versus Quantitative Predictions......Page 43
Quantitatively Predictive Logical Models......Page 46
Simplicity and Avoidance of Absurdity......Page 47
Ideal Gas Law and Engineering Freshmen......Page 49
4 Example of Model Building: Electoral Volatility......Page 52
Constructing a Coarse “Ignorance-Based” Model......Page 53
Testing the Coarse Model......Page 57
Refined Ignorance-Based Model......Page 58
How Precisely Can the Number of Parties Predict Volatility?......Page 60
The Main Contrasts Between Predictive and Descriptive Approaches......Page 62
Can Data with Low R² Confirm a Model?......Page 64
Some General Features of Constraint-Based Models......Page 66
Continuity......Page 67
Further Refinements and Aggregate Volatility......Page 68
5 Physicists Multiply, Social Scientists Add—Even When It Does Not Add Up......Page 70
Multiplication–Division Versus Addition–Subtraction......Page 72
Division: Widespread Versus Absent......Page 73
The Number of Variables per Equation: Occam’s Razor Versus Garbage Can......Page 74
The Number of Freely Adjustable Constants/Coefficients: Too Many Notes......Page 76
Alternate Equations for the Same Phenomenon: It is the Butler—Unless the Younger Uncle is Included......Page 79
Conceptual Consistency: Concern Versus Acceptance of Absurdity......Page 80
Prediction Versus Postdiction......Page 81
The Number of Decimals Reported: Meaningful Versus Meaningless......Page 82
Reversible and Transitive Versus Unidirectional and Nontransitive......Page 83
Interlocked Equations Versus Isolates......Page 84
Conclusions: Parsimony Versus Profusion......Page 85
An Actual Sequence in Physics—Basic Electricity......Page 87
6 All Hypotheses Are Not Created Equal......Page 89
Of Bones and Hypotheses......Page 90
The Dance Around the Null Hypothesis......Page 91
Directional Hypotheses......Page 92
Degrees of Falsifiablility and Usefulness......Page 93
False Positives: “p = .01” Does NOT Mean Confirmation in 99% of Replications......Page 95
The Tennis Match Between Data and Models......Page 96
Conclusions......Page 97
Devaluation of the Term “Model” and of Model Testing......Page 98
7 Why Most Numbers Published in Social Sciences Are Dead on Arrival......Page 100
Everyday and Scientific Thinking—and How Today’s Social Sciences Fit In......Page 101
A Crazy Methodology?......Page 102
Do Social Sciences Face a Ptolemaic Syndrome?......Page 104
The Meaning of Published Results: Steppingstones Versus Endpoints......Page 105
Does this Description Fit the Present Practices in Social Sciences?......Page 106
Conclusion: We Must Do Better than That......Page 109
Part II: Quantitatively Predictive Logical Models......Page 111
8 Forbidden Areas and Anchor Points......Page 113
No Forbidden Areas: Linear Relationships......Page 114
Only Positive Quadrant Allowed: Fixed Exponent Relationships......Page 115
Two Quadrants Allowed: Exponential Relationships......Page 118
Two Quadrants Partly Allowed: Logistic Relationships......Page 122
Constraints on Three Sides......Page 124
Constraints on All Four Sides......Page 125
Overview: Logically Predicted Forms, Empirically Determined Parameters......Page 128
Several Input Variables......Page 129
Conclusions: Why Would the Simplest Forms Prevail?......Page 132
Getting a Feel for Exponents and Logarithms......Page 134
9 Geometric Means and Lognormal Distributions......Page 138
Arithmetic Versus Geometric Means......Page 139
Normal Versus Lognormal Distributions......Page 142
Conclusions......Page 144
“Log-Lognormal” Distributions......Page 145
Mapping Any Limited Range on the Entire Range of Real Numbers......Page 146
“Less-than-Lognormal” Distributions......Page 147
From Assembly Size and District Magnitude to Mean Duration of Cabinets......Page 148
Why the Role of Assembly Size Can Emerge only from Logical Modeling......Page 152
Is It an Interlocking Network?......Page 153
All Predictive Models are not “Substantive”......Page 155
Conclusions......Page 156
11 Beyond Constraint-Based Models: Communication Channels and Growth Rates......Page 157
The Number of Communication Channels......Page 158
Inverse Square Law of Cabinet Duration......Page 159
Cube Root Law of Assembly Sizes......Page 160
Minimizing or Maximizing a Function......Page 161
Rate Equations......Page 162
Effective Number of Polities over 5,000 Years......Page 163
World Population Growth......Page 164
Trade/GNP Ratio......Page 165
Avoidance of Logical Inconsistencies: The Law of Minority Attrition......Page 166
The City–Country Rule: The Number of Components and the Number of Items in the Top Component......Page 168
Conclusions......Page 170
12 Why We Should Shift to Symmetric Regression......Page 172
The Standard OLS Regression Line Is Not a Trend Line......Page 173
Why Good Statistics Can Be Death of Predictive Science......Page 176
The Directionality of Standard OLS Equations......Page 178
Scale-Independent Symmetric Regression—A Single Line, with Transitivity......Page 180
Unemployment Versus Inflation Versus Unemployment: The Two Phillips Curves......Page 182
More than One Input Variable......Page 183
Connection to Scientific Practices......Page 184
Is Standard OLS Slope a Mixed Measure of Slope and Scatter?......Page 185
Three Interconnected Problems with Standard OLS......Page 187
The Mathematics of Symmetric Regression......Page 191
Give Preference to Indices That Best Fit a Logically Expected Relationship......Page 194
The Need for Symmetric Regression......Page 196
The Inverse Cube Root Relationship of Cabinet Duration and the Seat Product......Page 198
Dimensional Analysis......Page 199
Part III: Synthesis of Predictive and Descriptive Approaches......Page 203
14 From Descriptive to Predictive Approaches......Page 205
Could Statistical Approaches Find Social Laws?......Page 206
What Is Theory?......Page 210
Cycles and Sub-Cycles in Scientific Procedure: At what Stages Do Predictive and Descriptive Approaches Enter?......Page 213
Data: Graph It!......Page 217
Graph More than the Data!......Page 220
Consider Graphing by Hand......Page 222
Turning the Pattern Linear......Page 223
How to Publish Regression Results......Page 224
Conclusion: Ten Recommendations for Running and Reporting Linear Regression......Page 228
How NOT to Publish Regression Results......Page 229
16 Converting from Descriptive Analysis to Predictive Models......Page 233
Constructing a Predictive Model......Page 234
Guessing at the Values of Constants in the Model......Page 236
Inferring Predictive Model Constants from Regression Coefficients......Page 237
Conclusions......Page 240
Connection Between the Regression Coefficients in Samuels (2004)......Page 241
17 Are Electoral Studies a Rosetta Stone for Parts of Social Sciences?......Page 243
Interconnections of Scientific Disciplines......Page 244
Evolution of Quantitative Formalism......Page 246
Can Social Sciences Achieve the Degree of Predictive Ability of Natural Sciences, and Does It Matter?......Page 247
How Electoral Studies Look More Akin to Natural Sciences......Page 249
How Temperature Became a Ratio Variable......Page 250
How Electoral Studies Could Still Be a Rosetta Stone for Some Parts of Social Sciences......Page 252
Conclusions......Page 253
18 Beyond Regression: The Need for Predictive Models......Page 254
References......Page 259
Index......Page 267