Advancing the Frontiers of Simulation (eBook)

Preface 6
Contents 9
Contributors 11
George Fishman's Professional Career 14
1 A Brief Biography 14
2 Contributions to Computer Simulation 19
2.1 Contributions to Research 19
2.2 Dissemination of Knowledge and Advancement of the Field 24
2.3 Development of Software 26
3 George's Academic Family Tree 26
4 Recapitulation 28
References 29
A Conversation with George Fishman 33
1 Education and Early Career 33
2 Academic Career 41
3 Life After Being Department Chair 48
4 The Future of Simulation and Operations Research 51
Computer Intensive Statistical Model Building 55
1 Introduction 55
2 The Linear Model 57
3 Methods for Selecting the Best Model 59
3.1 “Min Cp” and “Unbiased Min p” Selection Methods 59
3.2 Dimensionality Problem 63
4 Bootstrap Analysis 63
4.1 Bootstrap Samples 63
4.2 Bootstrap Generation of a Set of Promising Models 64
4.3 Bootstrap Quality Assessment of Selected Best Model 66
5 Numerical Examples 67
5.1 Epoxide Bond Example 67
5.2 Bank Data Example 69
6 Conclusions 73
References 74
Patchwork Distributions 76
1 Introduction 76
2 Patchwork Distributions 78
3 Modeling Power 83
4 Modeling Effort: Theoretical Results 87
5 Modeling Effort: Computational Results 92
6 Conclusions 95
References 96
Asymptotic Validity of Batch Means Steady-State Confidence Intervals 98
1 Introduction 98
2 Validity Based on a FCLT Hypothesis 100
3 Validity Based on a CLT Hypothesis: Reductionto the Stationary Setting 102
4 Validity Based on a CLT Hypothesis: The Main Result 106
5 Validity Based on a CLT Hypothesis: Specializing the Results to the Regenerative Setting 109
References 115
Efficient Modeling of Delays in Discrete-Event Simulation 116
1 Introduction 116
2 The Four Forms of Delays 117
2.1 Scheduled Delays 118
2.2 State-Based Delays 118
2.3 Compound Delays 118
2.4 User-Managed Delays 119
3 Demands Imposed by World-View and Simulation Tool Architecture 119
3.1 Introduction 119
3.2 The Event-Scheduling World-View 119
3.3 The Transaction-Flow World-View 121
3.4 The SLX World-View 123
3.5 Ramifications of World-Views 123
4 Demands Imposed by Application Characteristics and Modeling Style 124
5 Modeling Time-Based Delays 125
5.1 Introduction 125
5.2 History 125
5.3 Required and Desired Properties of Event List Algorithms 125
5.4 An Event List Algorithm Is a Highly Specialized Form of Priority Queue 126
5.5 Event List Implementation Approaches 127
5.6 Measuring Event List Algorithm Performance 127
5.7 What Is the Best We Can Do? 128
5.8 Basics of Calendar Queues 129
5.9 H2, a Calendar Queue Algorithm under Development 130
5.10 Observations on H2 132
5.11 Priority Queue Algorithm Computational Costs 133
6 Modeling State-Based Delays 134
6.1 Introduction 134
6.2 The Objective 134
6.3 Run-Time Data Structures 135
6.4 Implementation Strategy 136
6.5 A Complex ``wait until'' Example 138
6.6 Compiling ``wait until'' Statements 140
6.7 Historical Perspectives 143
7 Compound Delays 145
8 User-Managed Delays 145
9 Examples of ``wait until'' Performance 148
10 Conclusions 150
References 151
Sampling from Linear Multivariate Densities 153
1 Introduction 153
2 Linear Densities over Point-symmetric Domains 154
2.1 An Extension 156
3 Improved Rejection 156
3.1 Concave Densities 157
3.2 General Densities 158
4 Applications 158
4.1 Concave Densities and Ahrens Method 158
4.2 Computational Experience 160
4.3 Importance Sampling 160
5 Conclusion 160
References 161
Factor Screening in Simulation Experiments: Review of Sequential Bifurcation 162
1 Introduction 162
2 SB's Competitors 164
2.1 Classic Two-Level Factorial Designs 164
2.2 Frequency Domain Experimentation (FDE) 165
2.3 Supersaturated Designs 165
2.4 Group-Screening Designs 165
3 Sequential Bifurcation 166
3.1 Outline of Simplest SB 166
3.2 Mathematical Details of Simplest SB 169
3.3 Case Study: Ericsson's Supply-Chain Simulation 171
3.4 SB with Two-Factor Interactions 172
4 Conclusions and Future Research 173
References 174
F2-Linear Random Number Generators 177
1 Introduction 177
2 F2-Linear Generators 178
2.1 General Framework 178
2.2 Jumping Ahead 181
2.3 Combined F2-Linear Generators 182
3 Quality Criteria 184
4 Lattice Structure in a Space of Formal Series 187
5 Specific Classes of Generators 191
5.1 The LFSR Generator 191
5.2 The GFSR, Twisted GFSR, and Mersenne Twister 193
5.3 The WELL RNGs 194
5.4 Xorshift Generators 195
5.5 Linear Recurrences in F2w 195
6 Speed and Performance in Statistical Tests 196
6.1 Speed Comparisons 196
6.2 Statistical Testing 198
7 Conclusion 198
References 199
Opportunities and Challenges in Health Care Simulation 202
1 Introduction 202
1.1 Similarities among Health Care Systems 202
1.2 Dissimilarities among Health Care Systems 204
2 Simulation in Health Care Systems 205
2.1 Discrete-Event Simulation 206
2.2 System Dynamics 207
2.3 A Taxonomy of Health Care Models 208
3 Some Illustrative Health Care Simulation Models 209
3.1 Level 1: Disease Models 210
3.2 Level 2: Operational Models of Health Care Units 214
3.3 Level 3: Strategic Whole-System Models 217
3.4 Commentary on Examples 219
4 Opportunities 220
5 Challenges 222
5.1 Project Management Challenges 223
5.2 Modeling Challenges 224
5.3 Challenges from the Health Care System 226
6 Overcoming the Challenges 227
6.1 Develop the Product 228
6.2 Grow Demand for Simulation 229
6.3 Work with the Customer 230
7 Conclusion 232
References 234
Future Trends in Distributed Simulationand Distributed Virtual Environments 237
1 Introduction 237
2 Survey on Future Trends in Distributed Simulationand Distributed Virtual Environments 238
2.1 Introduction 238
2.2 Survey Evaluation 240
3 Summary and Conclusions 258
References and Selected Bibliography 260
Combined Screening and Selection of the Best with Control Variates 268
1 Introduction 268
2 Generic Combined Procedure 269
3 Screening Procedure with Individual Control Variates 270
3.1 Individual Control-Variate Estimators 270
3.2 Screening with Individual Control Variates 272
4 Screening Procedures with Paired Control Variates 273
5 Selecting the Best with Control Variates 275
6 Control-Variate Selection for Screening 277
7 Combined Procedure 278
8 Empirical Results 280
8.1 Configurations and Experiment Design 281
8.2 Summary of Results 282
9 Illustration 283
10 Conclusions 284
References and Selected Bibliography 294
Optimal Linear Combinations of Overlapping Variance Estimators for Steady-State Simulation 295
1 Introduction 295
2 Basic Concepts and Assumptions for Steady-State Simulation Output Analysis 299
2.1 Nonoverlapping Batch Means (NBM) 300
2.2 Overlapping Batch Means (OBM) 300
2.3 Standardized Time Series (STS) 301
2.4 Key Asymptotic Properties of Selected Variance Estimators 303
3 Optimal Linear Combination of Overlapping Variance Estimators (OLCOVE) 304
3.1 Motivation 304
3.2 Covariance between Overlapping CvM Estimators 306
3.3 Covariance between OBM Estimators 310
3.4 General Linear Combinations of Overlapping Variance Estimators 314
3.5 Monte Carlo Study 317
4 Efficiency Analysis 319
5 OLCOVE Density and Confidence Intervals for µ and s2 321
5.1 Approximation to Asymptotic Density of an OLCOVE 321
5.2 Confidence Intervals for µ 324
5.3 Confidence Intervals for s2 327
6 Limitations and Evaluation of the OLCOVE Technique 329
7 Summary and Recommendations 329
References and Selected Bibliography 331
Author Index 333