ABSTRACT
Network analysis is a relatively novel approach to identify possible biomarker-related mechanisms in a disease being modulated by treatment. It is important to know how to interpret the visual representation of the network, as well as to control for overfitting the model. Partial correlation is a statistical tool used to perform network analysis of multivariate data. In psychology, latent class variables have been traditionally modeled via structural equation modeling for a long time. Regularization techniques like LASSO from regression models are available for graphical models as well. The R package qgraph integrates modeling, visualizing, and regularizing the partial correlation network in a comprehensive manner. The qgraph object can be stored in a variable and manipulated to produce the regularized correlation coefficients. The primary relationship between outcome and other covariates (first-order) are also detected by the regression model, which is contained in the object ‘reg’ for comparison.
