Multilevel Modeling for Longitudinal Analysis

D. Wayne Osgood
Pennsylvania State University

Abstract
This workshop focuses on how to address research questions about development and change by analyzing longitudinal data using multilevel regression analysis. Multilevel regression models are widely available through programs such as HLM, MLwiN, and Proc Mixed, but many scholars are unsure about how to apply them in their own work. The emphasis of the workshop will be on taking advantage of the flexibility of the multilevel framework to develop analyses that suit your research questions. Many researchers are now familiar with growth curve models, the version of multilevel models that has received the most attention, but they discover that those models are often a poor match to the problems they want to address. We will therefore spend most of our time on other uses of multilevel models in longitudinal research, such as analyzing discontinuous change over time, piece-wise regression, and time-varying explanatory variables. Through these tools we will be able to formulate models that flexibly characterize trajectories of development, assess whether variables account for both average development and individual differences in development, and gauge both the immediate and longer term effects of life events and interventions.

Biography
D. Wayne Osgood is a Professor in the Crime, Law and Justice Program of the Department of Sociology at Pennsylvania State University. He received his Ph.D. in social psychology from the University of Colorado, Boulder in 1977. He is a member of the MacArthur Research Network on Transitions to Adulthood and of the National Consortium on Violence Research, and he is a fellow of the American Society of Criminology. His research focuses on delinquency and other deviant behaviors of adolescence and early adulthood as well as quantitative methods for social research. He has published substantive research on peers and delinquency, the effectiveness of programs addressing delinquency and problem behavior, the transition to adulthood, time use and deviance, criminal careers, and the generality of deviance. His methodological articles have concerned multi-level models for longitudinal research and program evaluation, scaling self-reported delinquency, limited and discrete dependent variables, and Poisson-based analysis of aggregate data.