Latent variable models for assessing interaction effects in cognitive neuroscience

Abstract

The aim of this study is to provide latent variable models (LVM) to evaluate interaction effects between functional brain cortical regions. Rapidly developed imaging techniques and methods, for instance, ERP and fMRI, can serve as feasible accessories on mapping brain functions by monitoring activities among different cortical regions. Without capable statistical models, for example, the proposed latent variable models, analyzing these complex brain images can be troublesome. The advanced LVM proposed in this study can provide appropriate evaluations on interaction/connectivity among cortical regions that have drawn many attentions in recent cognitive neuroscience literature. We further demonstrate the feasibility of LVM by showing satisfying LVM accuracy rates under various simulated sample sizes and magnitudes of effects in evaluating cortical interactions. Practical suggestions and interpretations of this study are established to serve guidelines for medical and substantive researchers.