Master Class: Flexibly Using Structural Equation Modeling for Randomized Experiments

12 October 2023

Dr. Elisa Volpi, coordinatrice CUSO

Levente Littvay, Research Professor at the Centre for Social Sciences, Hungarian Research Network, and Academic Coordinator at MethodsNET.org

Summary: Structural Equation Models offers a flexible modeling approach for the social sciences (and beyond). In psychology and the rest of the social sciences, it is often discussed as the tool that allows for measurement error-free treatment of multiple indicator latent variables, and the testing cross-cultural invariance of survey instruments, it can do so much more than this. As randomized experiments become increasingly popular, multiple-indicator constructs often become outcomes in such studies. However, the status quo of how such models are estimated using structural equations tracks the logic of simple regressions and fails to take full advantage of structural equation models' full flexibility. An underutilized alternative approach built on multiple group analysis tools offers so much more. Prof. Littvay's talk will introduce the logic underlying how randomized experiments could be more flexibly analyzed using structural equation models.

Master Class: Beyond the specific topic of the talk, Prof. Littvay will consult projects using regression-based models, factor analyses, and structural equation models. This includes randomized and quasi-experimental designs and observational studies in the social sciences especially related to (but not limited to) political and voting behavior.

UNIGE

7