Comparative Analysis of Statistical Methods for Investigating Risk Factors of Adolescent Depression

Abstract

Depression often makes it difficult for individuals to form normal interpersonal relationships and may even lead to suicidal tendencies. Adolescents are in a critical period of developing psychological independence, with the prevalence of depression showing an increasing trend, making it a major global public health challenge. To achieve early identification and effective intervention, it is particularly important to employ appropriate statistical models to explore factors associated with adolescent depression. This article reviews the applications of multiple linear regression, logistic regression, linear mixed-effects models, and mediation models. The literature examines the specific problems addressed and the types of data suitable for each model in adolescent depression research. Based on the theoretical features of these models, comparisons are made to clarify how to choose the appropriate model under different research scenarios. Findings indicate that logistic regression is suitable for binary outcomes such as "whether an individual has depression"; multiple linear regression is often used for continuous variables such as depression scores; linear mixed-effects models are more appropriate for long-term follow-up studies of the same population or research involving individual variations; while mediation models are applicable when researchers aim to understand the mechanisms through which a certain factor influences the outcome.