Empirical Comparison of Bayesian Networks and Structural Equation Models in the Context of Educational Assessment

• In Event: Poster Session 9
  In Poster Session: Applications of Cognitive Diagnostic Models: From Q-Matrix Development to Data-Model Fit

Abstract

Making inferences about unobservable variables using directly observed data has been the primary interest of educational and psychological measurement. Two particular approaches, Bayesian networks (BNs) and Structure equation modeling (SEM), have become pervasive methods to serve as measurement models for educational assessment data. The present study discusses methodological similarities and differences, and then demonstrates how two approaches function differently using data set from a diagnostic assessment system called ACED. The results will guide researchers deciding which approach is most suitable for a given project when building cognitively diagnostic assessments.

Authors

• Yoon Jeon Kim, Worcester Polytechnic Institute

• Mengyao Cui, Florida State University

• Russell Almond, Florida State University