Psychometric Network Model Recovery: The Effect of Sample Size, Number of Items, and Number of Nodes

In recent years, network psychometrics has emerged as an alternative to the reflective latent variable model. This model conceptualizes traits as complex systems of behaviors mutually interacting with each other. Although this model offers important advantages compared to the reflective model, questions remain regarding the necessary sample size and the influence of factors such as the number of nodes and edges. This study aims to evaluate the psychometric network model performance under different conditions of sample size, number of nodes, and number of edges. The methodology involved a simulation with 1000 replicates for each combination of sample size, number of nodes, and the value of gamma parameter, which is used to determine the magnitude of the edges considered significant. The effect of these conditions on the accuracy of edge estimations and centrality indices (strength and expected influence) was assessed using sensitivity, specificity, and bias indicators. Results suggest that sample size and network complexity have a more significant impact than γ, methodological guidelines being proposed to support decision-making in applied research. In summary, this study provides empirically grounded recommendations that can guide applied researchers in designing robust psychometric network analyses and ensuring reliable estimation of model parameters.