Please use this identifier to cite or link to this item: https://ah.nccu.edu.tw/handle/140.119/99903


Title: 國中生人際衝突多層次潛在類別Mixture模式分析
Other Titles: Multilevel LCA Mixture Analysis of Interpersonal Conflicts among Middle School Students
Authors: 王郁琮;溫福星
Keywords: 多層次潛在類別分析;多層次Mixture模式;國中生人際衝突
multilevel latent class analysis;multilevel mixture model;interpersonal conflicts
Date: 2013-03
Issue Date: 2016-08-09 16:05:54 (UTC+8)
Abstract: 本研究利用多層次潛在類別無母數模式與多層次Mixture模式,針對國中生人際衝突二元資料同時進行學生及班級多階層類型探索,並對班級衝突脈絡變項進行多階層因素分析。研究樣本來自國中1~3年級85個班級共2,783人。學生層次分析顯示,國中生人際衝突族群分為三類,分別為「行為」、「語言」與「和諧」衝突族群。其中男生較易被歸為「行為」衝突;女生較易被歸為「語言」衝突。班級層次分析顯示,班級特定機率比值為單一類型而班級特定隨機指標建構出單因子模式,但個別因素負荷量不顯著,表示本研究在班級個數有限下,無法區分班級異質性。本研究進一步比較無母數與Mixture模式之分析結果,並針對人際衝突危機的實徵意涵與多層次潛在類別Mixture模式的技術應用進行詳細論述。
Conventional latent class analysis (LCA) classifies subjects into various categories by analyzing their response patterns to observed variables. However, multilevel latent class analysis (MLCA) is methodologically more appropriate when data are collected from a nested structure sample. Primary strength of MLCA is to analyze student and classroom levels of data simultaneously while taking the nested structure of the data into account. This study utilized and compared two different MLCA approaches including nonparametric MLCA and MLCA mixture models. Gender and proportion of boys in class are level 1 and level 2 covariates, respectively. Data were collected from 85 junior high classes with a total of 2,783 7th to 9th grade students. The observed variables include five binary self-reported survey questions regarding experiences related to interpersonal conflicts at school. All analyses were carried out by using Mplus6.0. Results show that non-parametric MLCA and MLCA mixture models fit the data equally well. Students are clustered into three categories namely "peaceful", "vocal" and "behavioral" of interpersonal conflicts. Boys are more likely to be classified as "behavioral" conflict while girls are more likely to be classified as "vocal" conflict. Nonparametric model clusters level 2 random intercepts as one class, whereas mixture model extracts the covariance among five level 2 random indicators as single factor possibly due to small number of level 2 units. Results of nonparametric and mixture MLCA were also discussed. Empirical implications as well as methodological challenges applying MLCA nonparametric and mixture models, are discussed in the end of the study.
Relation: 教育與心理研究, 36(1),89-115
Journal of Education & Psychology
Data Type: article
DOI 連結: http://dx.doi.org/10.3966/102498852013033601004
Appears in Collections:[教育與心理研究 TSSCI] 期刊論文

Files in This Item:

File Description SizeFormat
36(1)-89-116.pdf9338KbAdobe PDF611View/Open


All items in 學術集成 are protected by copyright, with all rights reserved.


社群 sharing