CYFS faculty affiliate Matt Fritz was invited to give a presentation at The Pennsylvania State University on March 23 as part of the Prevention and Methodology Training colloquia series. He shared a presentation titled “The mediated effect is significant, so I’m done, right? Thinking critically about testing and interpreting mediated effects in prevention interventions.”
Fritz discussed the need for prevention scientists to make more precise hypotheses regarding the shape and timing of the relations between variables in their intervention studies. He encouraged participants to look beyond statistical mediation analysis and to also consider study design.
While longitudinal models for statistical mediation have advanced in recent years, Fritz said, that is not always the case for the prevention intervention study designs to which these models are applied.
“The problem becomes that the statistical results are only valid when a large number of assumptions are met, almost all of which must be addressed in the design phase of the study—not the analysis phase,” said Fritz. “In the end, good statistics cannot make up for bad design.”
Fritz, whose research focuses on the development of statistical methods for modeling longitudinal mediation effects in prevention interventions, is an assistant professor in the Department of Educational Psychology and a core faculty member with the Nebraska Academy for Methodology, Analytics and Psychometrics.
Prevention and Methodology Training is an interdisciplinary program to train scientists in the integration of prevention science and statistical methodology. PAMT funded by the National Institute of Drug Abuse through a T32 grant to the Bennett Pierce Prevention Research Center and the Methodology Center.