A Primer on Regression Artifacts

by Campbell, Donald T.; Kenny, David A.

• ISBN13: 9781572308596
• ISBN10: 1572308591
• Edition: 1st
• Format: Paperback
• Copyright: 2002-12-11
• Publisher: The Guilford Press

Summary

Regression toward the mean is a complex statistical principle that plays a crucial role in any research involving the measurement of change. This primer is designed to help researchers more fully understand this phenomenon and avoid common errors in interpretation. The book presents new methods of graphing regression toward the mean, facilitating comprehension with a wealth of figures and diagrams. Special attention is given to applications related to program or treatment evaluation. Numerous concrete examples illustrate the ways researchers all too often attribute effects to an intervention or other causal variable without considering regression artifacts as an alternative explanation for change. Also discussed are instances when problems are actually created, instead of solved, by "correction" for regression toward the mean. Throughout, the authors strive to use nontechnical language and to keep simulations and formulas as accessible as possible.

Author Biography

Donald T. Campbell, Ph.D., before his death in 1996 was University Professor of Social Relations, Psychology, and Education at Lehigh University.

Table of Contents

Graphical Introduction	p. 1
From the Scatter Plot to the Correlation Coefficient	p. 3
The Pair-Link and Galton Squeeze Diagrams	p. 12
Backward Prediction	p. 16
Biology and Regression toward the Mean	p. 17
A Golden Oldie: McNemar's Illustration	p. 18
Conclusion	p. 18
Mathematical and Special Cases	p. 21
Generalization to the Case with Unequal Means and Variances	p. 21
Formal Definition and Formulas	p. 25
Frequently Asked Questions about Regression toward the Mean	p. 28
Conclusion	p. 35
Regression Artifacts Due to Extreme Group Selection	p. 37
Pre-Post Design and Regression Artifacts	p. 37
Examples	p. 40
Conclusion	p. 51
Regression Artifacts Due to Matching	p. 53
Detailed Illustration of the Limits of Matching	p. 55
Direction of Bias	p. 65
Conclusion	p. 66
Regression Artifacts Due to Statistical "Equating"	p. 68
Look at the "Treatment Effect" on the Covariate	p. 70
Illustration of the Bias in Statistical Equating	p. 71
Direction of Bias	p. 74
When Statistical Equating Works	p. 78
Alternatives to Statistical Equating	p. 79
Examples	p. 82
Conclusion	p. 84
Regression Artifacts in Change Scores	p. 87
Correlations with Change Scores	p. 88
Reliability of Change Scores	p. 90
The Measurement of Change	p. 91
Correlates of Change	p. 97
Conclusion	p. 98
Regression Artifacts in Time-Series Studies	p. 101
The Connecticut Crackdown on Speeding	p. 102
The Offset Effect of Psychotherapy	p. 105
Time Series of HIV Treatment in Clinical Trials	p. 110
ARIMA Modeling and Regression Artifacts	p. 116
Conclusion	p. 118
Regression Artifacts in Longitudinal Studies	p. 120
Over-Time Correlational Structure	p. 121
Regression Artifacts in Multiwave Studies	p. 131
Conclusion	p. 139
Cross-Lagged Panel Correlation Analysis	p. 141
What Is CLPC Analysis?	p. 141
Complications in CLPC Analysis	p. 146
The Rogosa Critique	p. 148
CLPC as a Special Case of the Multitrait-Multimethod Matrix Model	p. 151
Conclusion: What Is to Be Done?	p. 153
Conclusion	p. 155
Common Sense and Data Analysis: A Brief Epistemological Exegesis	p. 155
Time-Reversed Analysis	p. 158
Graphical Display of Results	p. 163
The Importance of Research Design	p. 164
Careful Consideration of Plausible Rival Hypotheses	p. 165
Regression and Prediction	p. 167
Conclusion	p. 170
Glossary of Terms	p. 173
Glossary of Symbols	p. 179
Dice-Rolling Program and Data Sets Used as Illustrations	p. 181
The Computation of Autocorrelations	p. 183
References	p. 185
Index	p. 195
Table of Contents provided by Rittenhouse. All Rights Reserved.