Conferences

ESRA2009:

Conference main page

Overview of sessions

Time table

Warsaw 2009: Presentations and short courses

Marginal causal models

Session: Marginal models for dependent data

Author:

Jacques Hagenaars; Tilburg University, Netherlands

Abstract:

Marginal modeling for dependent observations is often described as a ‘population averaged’ approach and as such opposed to causal modeling in the form of sturctural equation modeling (SEM), where the latter can be characterized as an individual-level or conditional approach. It will be shown that principles of marginal modeling enhance the usefulness of SEMs, for dependent but also for independent observations. This will be exemplified by applying SEMs (with latent variables) to categorical data from quasi-experimental designs (the pretest-posttest design and the nonequivalent group design) from family research and from multiple group comparisons where de observations in the groups are dependent.