Comparison of structural equation modeling and graphical modeling

Eötvös Loránd Tudományegyetem

Matematika Doktori Iskola

Rudas Tamás

matematika- és számítástudományok

These two contemporary approaches to modeling the association structure of multivariate data use graphs to visualize the models, but attribute different interpretations to the graphs. The former aims at modeling direct effects, the latter emphasizes conditional independencies. SEM was originally designed for Gaussian data but extensions to deal with categorical data are available, GM can handle normal and discrete variables, and their mixtures, equally well. Both methods offer the possibility of path analysis. The research will establish the identical and different characteristics of these two approaches and will compare their usefulness in the analysis of various data sets.

1 fő

Eötvös Loránd University

2017-05-31

english