The Issues of Construction and Analysis of Ordered Partial Correlation Networks in Psychological Research

Abstract

In the context of network modeling, psychological variables are understood as directly affecting each other, and not as a consequence of latent constructs. An efficient assessment of networks showing relationships between measured variables can be carried out using the methods of regularization of the network of partial correlations. This article provides an example of constructing a regularized network of partial correlations in the R software environment (it is showing the relationship between the personality traits of adolescents and their behavior in virtual space using the example of the social network VKontakte) and examines the features of constructing and analyzing ordered networks of partial correlations. A list of potential problems arising when using the considered network methodology is presented. The issues related to sample size and reproducibility of the network, difficulties in interpreting networks, and comparing different networks with each other, including both network models and models of latent variables, are considered.