All RECSM seminars are free of charge and no prior registration is necessary to attend them.
SEMINAR: "The future of social change models in social science"
(To view complete information on the seminar in .pdf format click here)
Date: Friday, March 8th, 2013, at 10.00.
Location: Universitat Pompeu Fabra, Ciutadella Campus. Ramon Turró Building, room 13.102. Carrer de Ramón Trias Fargas 25-27 (metro Ciutadella - Villa Olímpica). Barcelona 08005, Spain. (map).
Speakers: Dr. Manuel Voelkle (Center for Lifespan Psychology - Max Planck Institute for Human Development, Germany), Dr. Marc Delsing (Praktikon, Radboud University Nijmegen, The Netherlands), Dr. Han Oud (Behavioural Science Instutute - Radboud University Nijmegen, the Netherlands.
An easy introduction to continuous time models (Manuel Voelkle)
Most studies in the social sciences are concerned with outcome criteria that develop continuously over time. In practice, however, it is usually impossible to monitor the developmental process in a truly continuous manner. Rather, researchers are forced to use "snapshots" in order to learn something about the underlying process and factors that possibly affect it. Panel designs, in which the same subjects are repeatedly observed, are typical examples of such "snapshots". In this seminar, we first review several popular approaches to analyze panel data, including multilevel, autoregressive, and cross-lagged models. Advantages and limitations of each approach are discussed. Second, continuous time modeling is introduced as an alternative, which overcomes several limitations of the previous methods. To this end we make use of simulated and empirical data of a recently completed study. Third, we demonstrate how to estimate and interpret continuous time models using CT-SEM, an R-based function for continuous time modeling using SEM.
Continuous time analysis of causal directions between adolescents' internalizing and externalizing problems (Marc Delsing)
It is generally acknowledged that internalizing and externalizing problems are closely related and often co-occur. This comorbidity may result from various patterns of influence between internalizing and externalizing symptoms. The cross-lagged panel model and the latent growth curve model have become popular tools to assess the causal pathways between both types of problem behavior in a non-experimental context. Unfortunately, both methods have serious limitations that preclude causal conclusions. In this presentation, it is shown how these limitations can be circumvented by means of continuous time analysis of the cross-lagged panel design. The procedure is applied to the analysis of three waves of annually collected self-report data on adolescents' internalizing and externalizing problems.
Competitors and extensions of CT-SEM (Han Oud)
CT-SEM is evaluated as a stage in a series of longitudinal modeling efforts in social science. By far the most popular modeling approaches in social science have been the Autoregressive Cross-Lagged Panel and Latent Growth Curve models. Bollen and Curran combined basic features of both ("The best of two worlds") into the Autoregressive Latent Trajectory (ALT) model, for which Delsing and Oud formulated a CT version, called CALT model. The different approaches motivated McArdle to introduce the Latent Change Score (LCS) model. All will be related to CT-SEM and the advantages of CT-SEM explained. Finally, to apply a CT-SEM model in practice, an extension of the CT-SEM program in the form of a monitoring system will be discussed, which provides developmental curves and evaluates intervention measures for individual subjects in the sample.
About the speakers:
Dr. Manuel Voelkle is a research scientist at the Max Planck Institute for Human Development, Center for Life Span Psychology, in Berlin. He is particularly interested in the design and analysis of multivariate empirical studies with an emphasis on the use of structural equation models for the analysis of longitudinal data. Most of his current work is concerned with continuous time modeling and the analysis of the intricate relationship of between- and within-person differences in various constructs as they evolve over time.
Dr. Marc Delsing is a researcher at Praktikon Nijmegen. His research interests include adolescent problem behavior, the effectiveness of youth care interventions, youth culture, and longitudinal methodology. He is co-developer of BergOp, a web-based software program for monitoring treatment progress.
Dr. Han Oud is associate professor in longitudinal research methods and statistics at the Behavioural Science Institute of the Radboud University Nijmegen. He has taught psychometrics, general and longitudinal research methods and statistics. His research interests are in test construction, state space modeling, monitoring system construction, structural equation modeling (SEM), and longitudinal research. He currently concentrates on continuous time analysis by means of SEM and monitoring system construction on the basis of continuous time models. He has published numerous papers, book chapters and edited several books in these fields.
For more information on the seminar please contact us at firstname.lastname@example.org.