How do you analyze pretest-posttest data? Difference scores? Percent change scores? ANOVA? In medical, psychological, sociological, and educational studies, researchers often design experiments in which they collect baseline (pretest) data prior to randomization. However, they often find it difficult to decide which method of statistical analysis is most appropriate to use. Until now, consulting the available literature would prove a long and arduous task, with papers sparsely scattered throughout journals and textbook references few and far between.
Analysis of Pretest-Posttest Designs brings welcome relief from this conundrum. This one-stop reference - written specifically for researchers - answers the questions and helps clear the confusion about analyzing pretest-posttest data. Keeping derivations to a minimum and offering real life examples from a range of disciplines, the author gathers and elucidates the concepts and techniques most useful for studies incorporating baseline data.
Understand the pros and cons of different methods - ANOVA, ANCOVA, percent change, difference scores, and more
Learn to choose the most appropriate statistical test - Numerous Monte Carlo simulations compare the various tests and help you select the one best suited to your data
Tackle more difficult analyses - The extensive SAS code included saves you programming time and effort
Requiring just a basic background in statistics and experimental design, this book incorporates most, if not all of the reference material that deals with pretest-posttest data. If you use baseline data in your studies, Analysis of Pretest-Posttest Designs will save you time, increase your understanding, and ultimately improve the interpretation and analysis of your data.
Table of Contents
Clinical Applications of Pretest-Posttest Data
Why use the Pretest Data
Graphical Presentation of Pretest-Posttest Data
How to Analyze Pretest-Postest Data: Possible Solutions
A Note on SAS Notation
Focus of the Book
What is Validity?
What is Reliability?
What is Regression Towards the Mean?
Why is Regression Towards the Mean Important?
Dealing with Regression Towards the Mean and How to Take Advantage of Test-Retest Reliability
What is Pretest Sensitization?
Controlling for Pretest Sensitization with Factorial Designs
Alternative Methods for Controlling for Pretest Sensitization
Definition and Assumptions
Case 1: The Absence of a Treatment Intervention Between Measurement of the Pretest and Posttest Scores
Case 2: The Application of a Treatment Intervention Between Measurement of the Pretest and Posttest Scores
Nonparametric Alternative to Case 1 or Case 2
Case 3: Two Groups with Different Treatment Interventions Between Measurement of Pretest and Posttest Scores
Case 4: More than Two Groups with Different Treatment Interventions Between Measurement of Pretest and Posttest Scores
Unreliability of Difference Scores
Testing the Distribution of Change and Relative change Scores
Effect of Regression Towards the Mean on Difference Scores
RELATIVE CHANGE FUNCTIONS
Definitions and Assumptions
Statistical Analyses with Change Scores
Change Scores and Regression Towards the Mean
Difference Scores or Relative change Scores?
Other Relative change Functions
Distribution of Relative change Scores
ANALYSIS OF COVARIANCE
Definitions and Assumptions
ANCOVA with Difference Scores as the Dependent Variable
ANCOVA using Percent change as the Dependent Variable
Assumptions of the ANCOVA
Violation of Homogeneity of Within-Groups Regression Coefficients
Effect of Outliers and Influential Observations
Nonrandom Assignment of Subject to Treatment Groups
Using Stratification to Control for the Pretest
REPEATED MEASURES ANALYSIS OF VARIANCE
Using Repeated Measures ANOVA for Analysis of Pretest-Posttest Data
Regression Towards the Mean with Multiple Posttest Measurements
Using Repeated Measures ANOVA for Analysis of Pretest-Posttest Data with Multiple Posttest Measurements
Analysis of Repeated Measures using Summary Measures
CHOOSING A STATISTICAL TEST
Choosing a Test Based on how the Data will be Presented
Generation of Bivariate, Normally Distributed Data with a Specified Covariance Structure
Monte Carlo Simulation when the Assumptions of the Statistical Test are Met
Monte Carlo simulation when Systematic Bias Affects the Pretest and Posttest Equally
Monte Carlo Simulation when the variance of the Posttest Scores does not Equal the Variance of the Pretest Scores
Monte Carlo Simulation when Subjects are Grouped A Priori based on Pretest Score
Monte Carlo Simulation when the Marginal Distribution of the Pretest and Posttest Scores is Non-Normal
Permutation Tests and Randomization Tests
Randomization Tests and Pretest-Posttest Data
Analysis of Covariance
Resampling within Block or Time Periods
Resampling with Missing Data
SPECIAL TOPICS: EQUALITY OF VARIANCE
Methods and Procedures
APPENDIX: SAS Code