Week 7

Keanan Joyner - Assistant Professor, University of California, Berkeley

Leveraging Multilevel Structural Equation Modeling for Probing Individual Differences in Behavioral Performance and Neural Trial-Level Analyses

Multilevel Structural Equation Modeling (MSEM) is a flexible analytic tool combining advantages of both multilevel modeling and structural equation modeling. MSEM is gaining popularity for analyzing intensive longitudinal data, often collected via ecological momentary assessment. However, the trial structure of most cognitive neuroscience-style tasks would benefit from the analytic advantages of MSEM. This workshop focuses on trial-level analyses of behavioral performance and event-related potential responses to illustrate the utility of MSEM for these data types. First, introductory aspects of structural equation modeling and multilevel modeling separately will be briefly covered, and then advantages of the MSEM framework will be discussed. Then, an empirical example of application of MSEM to real data from an Eriksen Flanker task will be shown, revealing unique relationships with substance use constructs. Sample code will be made available for the models being fit in Mplus, and for a subset of models, in R using the ‘lavaan’ package. Trial-level analyses in cognitive neuroscience tasks can yield useful insights into risk and resilience factors implicated in a variety of psychological outcomes, and MSEM is a useful, flexible analytic tool to assist in this goal.