| Volume 2, Issue 2, 139-169, 2011 |
|Return to Issue List |
|How to conduct and statistically analyze case-based time series studies, one patient at a time|
|Michael R. Nash, Ph.D. (a), Jeffrey J. Borckardt, Ph.D. (b), Amineh Abbasa (a), Erin Gray (a)|
|(a) University of Tennessee|
(b) Medical University of South Carolina
|Volume 2, Issue 2, 2011, Pages 139-169|
|We describe how to conduct case-based time-series studies in a practice setting. First we offer a sampler of clinical research questions that can be addressed by case-based studies. Second we construct a hypothetical case that illustrates the structure of a time-series project now being conducted in a university-based outpatient psychotherapy clinic. This case also familiarizes the reader with the data array of a time-series study. Third, we present two actual case studies, each carried out in a different outpatient setting. Fourth, we move to the logistics of how a time-series study is efficiently conducted in an applied setting. Finally we provide a step-by-step description of Simulation Modeling Analysis (SMA) for time-series data and how the practitioner can use freely-available software to analyze his or her real-world clinical practice data (i.e., relatively short streams of time-series data). |
|Table of Contents|
I. The domain of clinical research questions addressed by case-based time-series studies
Questions of Improvement: Looking for an Effect for Phase
Time-series improvement studies bridge the gap between practice and the laboratory.
Questions of Change-Process: Looking for Patterns of Change During Treatment
How Time-series Process-Change Studies Help Bridge the Gap Between Practice and the Laboratory.
II. A real practice-based time-series project illustrated by an imaginary case
What is Autocorrelation?
How do we calculate Autocorrelation?
Why is Autocorrelation Important?
Applying SMA to the Hypothetical Case
III. REAL CASES
Case 1: An Improvement Study in an EAP Setting: Looking for an Effect of Phase.
Case 2: Private-practice setting: looking for patterns of change during treatment.
Process-Change: The Cross-Lag Correlation.
Time, Cause, inference, and Aristotle.
IV LOGISTICS: THE FUNDAMENTALS
Creating a Data Stream
Source and Content of the Observations.
Sample Evenly and Frequently.
Supplementing with a Standardized Outcome Measure.
Maintaining the Datastream.
V ANALYSIS OF TIME-SERIES DATA
Data Fluctuation in Group Designs
Data Fluctuation in Time-series Studies
Autocorrelation as Nuisance and Raison d'Etre
Simulation Modeling Analysis for Time-Series (SMA: Step by Step)
Appendix: Comparison of SMA to ITSACORR and Explanation of Simulation Modeling Calculations Using an Example Data Set
Simulation Modeling Explained with an Example:
|Michael R. Nash, Professor of Psychology, Psychology Department, 307 Austin Peay Building, University of Tennessee, Knoxville, TN 37996-0900. |
|time-series, statistics psychotherapy outcome |
|Received 26 Jul 2010; Revised 6 Dec 2010; Accepted 14 Dec 2010; In Press 5 May 2011 |