Published: April 10, 2020

There recently has been growing interest in the study of psychological and neurological processes at an individual level. One goal such endeavors is to construct person-specific dynamic assessments using time series techniques as Vector Autoregressive (VAR) models. However, two problems exist with current VAR specifications: 1) models are restricted that contemporaneous relations typically modeled either undirected among residuals or directed observed variables, but not both; 2) estimation frameworks limited by reliance on stepwise model building procedures. This adopts a new modeling approach. We first extended unified SEM (uSEM) framework, widely used structural model, hybrid representation (i.e., “huSEM”) include both effects, then replaced LASSO-type regularization for global search optimal sparse model. Our simulation showed regularized huSEM performed uniformly best over alternative representations and/or approaches, respect accurately recovering presence directionality reliably removing false when data generated have types relations. The present our knowledge application developed technique huSEM, which points promising future statistical learning psychometric

Language: Английский