Symptom Structure in Schizophrenia: Implications of Latent Variable Modeling vs Network Analysis

The structure of schizophrenia symptoms has a substantial impact on the development of pharmacological and psychosocial interventions. Typically, reflective latent variable models (eg, confirmatory factor analysis) or formative latent variable models (eg, principal component analysis) have been used to examine the structure of schizophrenia symptoms. More recently, network analysis is appearing as a method to examine symptom structure. However, latent variable modeling and network analysis results can lead to different inferences about the nature of symptoms. Given the critical role of correctly identifying symptom structure in schizophrenia treatment and research, we present an introduction to latent variable modeling and network analysis, along with their distinctions and implications for examining the structure of schizophrenia symptoms. We also provide a simulation demonstration highlighting the statistical equivalence between these models and the subsequent importance of an a priori rationale that should help guide model selection.