APRIL 18-21, 2017
Development of a Visual Analytics System for Structural Equation Modeling
In this study, we introduce a visual analytics system for structural equation modeling (SEM) that allows analysts to efficiently build and validate statistical models. Discovery of causality is one of the essential objectives of science. SEM is a statistical analysis method that can validate causal relationship model composed of observed and latent variables. SEM is widely used to validate causal relationship in complex phenomena and several support tools have been developed. However, existing support tools lack functionalities for integrated visualization of model information. Our proposed system simultaneously visualizes model information using 3 types of views: graph view, matrix view, and fitness measurement table view. We demonstrate an application example using real world data to validate the concept of our proposed system.