潜在剖面模型的后续分析——比较分类分析法改进后的偏差

发布时间：2018-03-07 06:30

  本文选题：潜在类别分析　切入点：潜在剖面模型　出处：《心理学探新》2017年05期 　论文类型：期刊论文

【摘要】：传统的分类分析法虽然是潜在类别模型常用的后续分析方法,但容易导致后续模型中潜在类别与其他变量之间关系的低估。现阶段已发展出多种改进的方法:一步法、基于模型的方法、Bartlett法,改进的分类分析法(包括ML三步法、BCH法、纳入式分类分析法)。本文对这些方法研究进行综述总结,进一步针对心理学研究数据的特点,使用模拟实验探讨适用于潜在剖面模型的分类分析方法,结果发现:传统方法低估潜在类别变量与因变量的关系;ML三步法只有在潜在类别概率分布平均时估计精确;BCH法估计最接近真值,但在低分类区分度、大效果量时出现概率估计为负值的情况;纳入法虽有轻微的高估,但在各种模拟条件下参数估计最为稳健。这些方法受分类区分度、类别概率均匀性以及潜在类别变量与附属变量关系的效果量所影响。
[Abstract]:Although the traditional classification analysis method is a common follow-up analysis method in the potential category model, it is easy to lead to the underestimation of the relationship between the potential category and other variables in the follow-up model. At present, many improved methods have been developed: one step method. The model-based method is Bartlett's method, the improved classification analysis method (including ML three-step method / BCH method), the inclusive classification analysis method, etc. In this paper, the research of these methods is summarized, and the characteristics of psychological research data are further discussed. Using simulation experiments to explore classification and analysis methods suitable for potential profile models, The results show that the traditional method underestimates the relationship between the potential class variables and the dependent variables. The ML three-step method only estimates accurately the true value when the probability distribution of the potential category is average, but the classification degree is low. The probability is estimated to be negative when the large effect is large. Although the inclusion method is slightly overestimated, the parameter estimation is the most robust under various simulation conditions. Class probability uniformity and the effect of the relationship between potential class variables and dependent variables.
【作者单位】： 深圳大学心理与社会学院;华南师范大学心理学院心理应用研究中心;
【基金】：国家自然科学基金项目(31700982) 深圳大学人文社会科学项目(85201-00000517) 广州市教育科学“十二五”规划2014年度重大课题(1201411413) 2016年广州市中小学教育质量阳光评价项目第二期(GZJY2051S/YD16G0510)
【分类号】：B841.2
，

本文编号：1578318