Within multidisciplinary fields like Decision Neuroscience multilevel datasets are increasingly common. For example, trial-level data which include behavioral variables (e.g., choice, uncertainty, difficulty, prior choices or outcomes, etc) and physiological variables (e.g., BOLD signal, heart rate, cortisol, etc) may be nested within blocks or treatments which may be nested within person-level individual difference variables (e.g., genotype, cognitive ability, age, etc). The data are often collapsed across some level or condition (sacrificing potentially interesting variance) to analyze these mixed model designs with more common techniques such as a repeated measures ANOVA. However, multilevel techniques that acknowledge the nested structure of the data are more appropriate and flexible for these integrative datasets. The first official training activity sponsored by the Network was focused on multilevel modeling and multilevel mediation and was hosted at the Kellogg School of Management at Northwestern University. The workshop occurred on the day before the 2011 Society for Neuroeconomics annual meeting.
This one-day workshop provided an introduction to multilevel modeling (MLM). The morning session was led by Shevaun Neupert who introduced MLM and discussed when, why, and how to use these methods. The afternoon session was led by Lauren Atlas who discussed multilevel mediation analysis of neuroimaging data.
Participants included graduate students, post-docs, research staff, and junior faculty. Travel costs were subsidized by the Scientific Research Network on Decision Neuroscience and Aging. This fund covered an airfare subsidy up to $300 and one night of lodging at the Hotel Orrington in Evanston.
Workshop program (PDF)
Introduction (PDF) G. Samanez-Larkin
Morning Session (PDF) S. Neupert
Afternoon Session (PDF) L. Atlas
About the Instructors
Shevaun Neupert is an Associate Professor of Psychology at North Carolina State University. She received her PhD from the University of Arizona. Her research focuses on daily stressors and their associations with affect, physical health, and memory across the life span, socioeconomic disparities in health, and statistical techniques for examining change and intraindividual variability. She teaches a graduate level course on multilevel modeling and co-leads an annual summer workshop on regression, latent variable modeling, and multilevel modeling.
Lauren Atlas is a post-doctoral fellow at NYU. She received her PhD from Columbia University where she was trained by methodologists Tor Wager, Martin Lindquist, and Niall Bolger. Her research focuses on the mechanisms by which expectancies modulate affective experience. Ongoing projects use a thermal pain model to characterize the brain pathways that are responsible for the transfer from objective stimulation (i.e. applied temperature) to subjective experience (i.e. perceived pain). This work uses fMRI alongside other methodologies, including psychophysiology, pharmacological interventions, and TMS, and takes advantage of advanced fMRI analysis approaches including whole-brain multi-level mediation analysis to identify the brain processes that formally link experimental manipulations with ongoing behavioral responses.