Design of Observational Studies (eBook)

Paul R. Rosenbaum (Autor)

eBook Download: PDF

2009 | 2010
XVIII, 384 Seiten
Springer New York (Verlag)
978-1-4419-1213-8 (ISBN)

Infos

An observational study is an empiric investigation of effects caused by treatments when randomized experimentation is unethical or infeasible. Observational studies are common in most fields that study the effects of treatments on people, including medicine, economics, epidemiology, education, psychology, political science and sociology. The quality and strength of evidence provided by an observational study is determined largely by its design. Design of Observational Studies is both an introduction to statistical inference in observational studies and a detailed discussion of the principles that guide the design of observational studies.

Design of Observational Studies is divided into four parts. Chapters 2, 3, and 5 of Part I cover concisely, in about one hundred pages, many of the ideas discussed in Rosenbaum's Observational Studies (also published by Springer) but in a less technical fashion. Part II discusses the practical aspects of using propensity scores and other tools to create a matched comparison that balances many covariates. Part II includes a chapter on matching in R. In Part III, the concept of design sensitivity is used to appraise the relative ability of competing designs to distinguish treatment effects from biases due to unmeasured covariates. Part IV discusses planning the analysis of an observational study, with particular reference to Sir Ronald Fisher's striking advice for observational studies, 'make your theories elaborate.'

The second edition of his book, Observational Studies, was published by Springer in 2002.

---

An observational study is an empiric investigation of effects caused by treatments when randomized experimentation is unethical or infeasible. Observational studies are common in most fields that study the effects of treatments on people, including medicine, economics, epidemiology, education, psychology, political science and sociology. The quality and strength of evidence provided by an observational study is determined largely by its design. Design of Observational Studies is both an introduction to statistical inference in observational studies and a detailed discussion of the principles that guide the design of observational studies.Design of Observational Studies is divided into four parts. Chapters 2, 3, and 5 of Part I cover concisely, in about one hundred pages, many of the ideas discussed in Rosenbaum's Observational Studies (also published by Springer) but in a less technical fashion. Part II discusses the practical aspects of using propensity scores and other tools to create a matched comparison that balances many covariates. Part II includes a chapter on matching in R. In Part III, the concept of design sensitivity is used to appraise the relative ability of competing designs to distinguish treatment effects from biases due to unmeasured covariates. Part IV discusses planning the analysis of an observational study, with particular reference to Sir Ronald Fisher's striking advice for observational studies, "e;make your theories elaborate."e;The second edition of his book, Observational Studies, was published by Springer in 2002.

Inhalt

Preface 7
Acknowledgments 8
Contents 9
Part I: Beginnings 15
1 Dilemmas and Craftsmanship 16
1.1 Those Confounded Vitamins 16
1.2 Cochran’s Basic Advice 17
1.3 Maimonides’ Rule 20
1.4 Seat Belts in Car Crashes 22
1.5 Money for College 23
1.6 Nature’s ‘Natural Experiment’ 24
1.7 What This Book Is About 26
1.8 Further Reading 31
References 31
2 Causal Inference in Randomized Experiments 34
2.1 Two Versions of the National Supported Work Experiment 34
2.2 Treatment Effects in Randomized Experiments 38
2.3 Testing the Null Hypothesis of No Treatment Effect 42
2.4 Testing Other Hypotheses Confidence Intervals
2.5 Attributable Effects 62
2.6 Internal and External Validity 69
2.7 Summary 70
2.8 Further Reading 70
2.9 Appendix: Randomization Distribution of m-statistics 71
References 74
3 Two Simple Models for Observational Studies 77
3.1 The Population Before Matching 77
3.2 The Ideal Matching 78
3.3 A Na ¨ ive Model: People Who Look Comparable Are Comparable 82
3.4 Sensitivity Analysis: People Who Look Comparable May Differ 88
3.5 Welding Fumes and DNA Damage 91
3.6 Bias Due to Incomplete Matching 97
3.7 Summary 98
3.8 Further Reading 99
3.9 Appendix: Exact Computations for Sensitivity Analysis 100
References 102
4 Competing Theories Structure Design 107
4.1 How Stones Fall 107
4.2 The Permanent-Debt Hypothesis 110
4.3 Guns and Misdemeanors 112
4.4 The Dutch Famine of 1944–1945 112
4.5 Replicating Effects and Biases 113
4.6 Reasons for Effects 116
4.7 The Drive for System 120
4.8 Further Reading 121
References 122
5 Opportunities, Devices, and Instruments 125
5.1 Opportunities 125
5.2 Devices 128
5.3 Instruments 143
5.4 Summary 152
5.5 Further Reading 152
References 153
6 Transparency 158
References 160
Part II: Matching 161
7 A Matched Observational Study 162
7.1 Is More Chemotherapy More Effective? 162
7.2 Matching for Observed Covariates 163
7.3 Outcomes in Matched Pairs 166
7.4 Summary 168
7.5 Further Reading 170
References 170
8 Basic Tools of Multivariate Matching 171
8.1 A Small Example 171
8.2 Propensity Score 173
8.3 Distance Matrices 176
8.4 Optimal Pair Matching 180
8.5 Optimal Matching with Multiple Controls 183
8.6 Optimal Full Matching 187
8.7 Efficiency 191
8.8 Summary 192
8.9 Further Reading 192
References 193
9 Various Practical Issues in Matching 195
9.1 Checking Covariate Balance 195
9.2 Almost Exact Matching 198
9.3 Exact Matching 200
9.4 Missing Covariate Values 201
9.5 Further Reading 202
References 202
10 Fine Balance 204
10.1 What Is Fine Balance? 204
10.2 Constructing an Exactly Balanced Control Group 205
10.3 Controlling Imbalance When Exact Balance Is Not Feasible 208
10.4 Fine Balance and Exact Matching 210
10.5 Further Reading 211
References 211
11 Matching Without Groups 213
11.1 Matching Without Groups: Nonbipartite Matching 213
11.2 Some Practical Aspects of Matching Without Groups 217
11.3 Matching with Doses and Two Control Groups 219
11.4 Further Reading 226
References 226
12 Risk-Set Matching 228
12.1 Does Cardiac Transplantation Prolong Life? 228
12.2 Risk-Set Matching in a Study of Surgery for Interstitial Cystitis 229
12.3 Maturity at Discharge from a Neonatal Intensive Care Unit 233
12.4 Joining a Gang at Age 14 236
12.5 Some Theory 237
12.6 Further Reading 238
References 239
13 Matching in R 241
13.1 241
13.2 Data 242
13.3 Propensity Score 244
13.4 Covariates with Missing Values 244
13.5 Distance Matrix 246
13.6 Constructing the Match 247
13.7 Checking Covariate Balance 248
13.8 College Outcomes 250
13.9 Further Reading 251
13.10 Appendix: A Brief Introduction to 252
13.11 Appendix: 254
Functions for Distance Matrices 254
References 256
Part III: Design Sensitivity 258
14 The Power of a Sensitivity Analysis and Its Limit 259
14.1 The Power of a Test in a Randomized Experiment 259
14.2 Power of a Sensitivity Analysis in an Observational Study 267
14.3 Design Sensitivity 271
14.4 Summary 274
14.5 Further Reading 274
Appendix: Technical Remarks and Proof of Proposition 14.1 274
References 276
15 Heterogeneity and Causality 277
15.1 J.S. Mill and R.A. Fisher: Reducing Heterogeneity or Introducing Random Assignment 277
15.2 A Larger, More Heterogeneous Study Versus a Smaller, Less Heterogeneous Study 279
15.3 Heterogeneity and the Sensitivity of Point Estimates 283
15.4 Examples of Efforts to Reduce Heterogeneity 284
15.5 Summary 286
15.6 Further Reading 286
References 286
16 Uncommon but Dramatic Responses to Treatment 288
16.1 Large Effects, Now and Then 288
16.2 Two Examples 291
16.3 Properties of a Paired Version of Salsburg’s Model 293
16.4 Design Sensitivity for Uncommon but Dramatic Effects 295
16.5 Summary 297
16.6 Further Reading 298
16.7 Appendix: Sketch of the Proof of Proposition 16.1 298
References 299
17 Anticipated Patterns of Response 300
17.1 Using Design Sensitivity to Evaluate Devices 300
17.2 Coherence 300
17.3 Doses 304
17.4 Example: Maimonides’ Rule 309
17.5 Further Reading 310
17.6 Appendix: Proof of Proposition 17.1 310
References 311
Part IV: Planning Analysis 313
18 After Matching, Before Analysis 314
18.1 Split Samples and Design Sensitivity 314
18.2 Are Analytic Adjustments Feasible? 316
18.3 Matching and Thick Description 321
18.4 Further Reading 323
References 323
19 Planning the Analysis 326
19.1 Plans Enable 326
19.2 Elaborate Theories 328
19.3 Three Simple Plans with Two Control Groups 329
19.4 Sensitivity Analysis for Two Outcomes and Coherence 338
19.5 Sensitivity Analysis for Tests of Equivalence 340
19.6 Sensitivity Analysis for Equivalence and Difference 342
19.7 Summary 344
19.8 Further Reading 344
19.9 Appendix: Testing Hypotheses in Order 345
References 349
Summary: Key Elements of Design 351
Solutions to Common Problems 353
References 356
Symbols 357
Acronyms 359
Glossary of Statistical Terms 361
Some Books 367
References 367
Suggested Readings for a Course 369
References 369
Index 371

Produkt-Details

• Erscheint lt. Verlag: 22.10.2009
• Reihe/Serie: Springer Series in Statistics
• Zusatzinfo: XVIII, 384 p.
• Verlagsort: New York
• Sprache: englisch
• Themenwelt: Geisteswissenschaften
• Themenwelt: Mathematik / Informatik ► Mathematik ► Statistik
• Themenwelt: Mathematik / Informatik ► Mathematik ► Wahrscheinlichkeit / Kombinatorik
• Themenwelt: Sozialwissenschaften ► Politik / Verwaltung
• Themenwelt: Technik
• Themenwelt: Wirtschaft ► Volkswirtschaftslehre ► Ökonometrie
• Schlagworte: Analysis • causal inference • natural experiment • observational study • propensity score • Sensitivity Analysis • Statistical Inference
• ISBN-10: 1-4419-1213-4 / 1441912134
• ISBN-13: 978-1-4419-1213-8 / 9781441912138