Title: Functional structural equation modeling with high dimensional data in medical research
Authors: Yuko Araki - Shizuoka University (Japan) [presenting]
Abstract: Structural equation modeling is a widely used multivariate statistical method to reveal structural relationships of several variables. In recent years in medical research, imaging data such as Magnetic Resonance imaging, have been more sophisticated and it its known that they give us important information. We propose a functional structural equation modeling to analyze association among several biomarkers including high dimensional image data. For effective and stable parameter estimation, the high dimensional data are expanded by using basis functions and sparse principal component analysis. We also discuss model selection methods based on information criterion. The proposed method is investigated through Monte Carlo simulations and applied to a medical study.