本研究旨在應用多層次混合迴歸模式（Multi-level Regression Mixture Model, MLR-Mixture Model），探討影響國中學生憂鬱情緒之因素，並與階層線性模式（HLM）作比較。研究變項分個人與團體層次：包含性別、霸凌危機、班級男女比例及班級霸凌氣氛等。藉由置入個體層次潛在類別變項，進而了解國中生憂鬱情緒之異質性分配來源。研究樣本共2,987名七～八年級學生，資料蒐集工具包含被霸凌危機，以及個人憂鬱情緒題項。MLR-Mixture Model分析結果發現，主觀憂鬱情緒呈現低、中、高三種類別，而學生性別、霸凌危機對學生憂鬱情緒具有顯著解釋力。就班級層次而言，班級霸凌氣氛對班級憂鬱氣氛也具解釋力，但班級男女比例則對班級憂鬱氣氛無顯著效果。文末針對MLR-Mixture Model的應用分析技術提出若干建議，也針對國中學生霸凌與憂鬱等相關議題作出剖析。

Understanding gender difference between crisis of being bullied and level of depression has been a major task for school counselors and educators. The purpose of this study was to apply the newly developed Multi-Level Regression Mixture Model (MLR-Mixture) to analyze latent heterogeneity. Variables at the individual level studied include gender and self-perceived crisis of being bullied and depression, and proportion of male in the class and mean score of bully were included as group level predictors. Three Mixture Models were tested and the results were compared with those of HLM. Current study includes 2,987 students of grades 7~9 and data were collected using two sets of binary self-report survey questions. Results show that sense of depression can be categorized into three latent groups namely low, middle and high depression group, respectively. Furthermore, depression was significantly influenced by gender, crisis of bully and collective sense of bully in the class but not by proportion of male. In addition, when heterogeneous subpopulations exist, using HLM for data analysis produces results that are completely misaligned with MLR-Mixture Model. Comments and suggestions were provided at the end for future studies.