Title: Fitting Like a Glove: A New Method for Constructing Networks for
Psychometric Data

Authors: Claudia D. van Borkulo, Denny Borsboom, Sacha Epskamp, Tessa F.
Blanken, Lynn Boschloo, Robert A. Schoevers, Lourens J. Waldorp

Abstract: In network approaches to psychopathology, psychological constructs
(e.g., depression) are viewed as complex systems of interacting entities
(e.g., interactions between symptoms like insomnia and fatigue). A crucial
step in the application of such network models lies in the assessment of
network structure, as coded in an adjacency matrix; in network models for
psychopathology, for example, this matrix specifies which symptoms directly
interact with each other and which do not. In the present paper, we present a
computationally efficient method for assessing network structures, which is
tailored to psychometric applications. The approach, which is based on Ising
models as used in physics, combines l_1-regularized logistic regression with
model selection based on the Extended Bayesian Information Criterion (EBIC) to
identify relevant symptom-symptom relationships that define connections in a
network. A validation study shows that this method succeeds in revealing the
most relevant features of a network for realistic sample sizes. Possible
extensions of the model are discussed.