Statistical Modeling of Behavior Change in Small Samples
Categories: Press Releases
Discovering the core components, processes, and mechanisms of intervention effectiveness is a primary objective in contemporary treatment development and dissemination research. Studies along these lines typically use complex multivariate statistical models to clarify clinically intricate associations among sets of variables hypothesized to produce clinical change. Unfortunately, implementing and executing multivariate statistical analyses requires large sample sizes which typically are infeasible within the budgets of most federally funded clinical trials.
To address the analytic obstacles posed by small sample sizes in multivariate clinical research, ORI scientist Tim Ozechowski, Ph.D., has written an article demonstrating the use of empirical Bayes Markov Chain Monte Carlo (EB MCMC) estimation for statistically analyzing linkages between in-session clinical processes, mechanisms of change, and behavioral outcomes using structural equation modeling with prohibitively small samples
Dr. Ozechowski’s paper provides a demonstration and tutorial on the use of EB MCMC estimation for clinical researchers unfamiliar with Bayesian statistical methods. The article will appear in a forthcoming special issue of the Journal of Consulting and Clinical Psychology on data analytic methods for evaluating treatment outcome and mechanisms of change.
“This paper provides clinical researchers with a sound statistical methodology for modeling clinically intricate linkages between in-session therapy processes, mechanisms of change, and behavioral outcomes when only a small sample is available, as is the norm in clinical research," noted Ozechowski.
This research was funded by National Institutes of Health grants R03DA017181 and R03DA021221 awarded to Timothy J. Ozechowski and R01DA09422 awarded to Holly B. Waldron.