概化理论两侧面设计方差分量及其变异量估计方法比较  

Comparison of Methods for Estimating Variance Components and Their Variabilities in Generalizability Theory Based on Two-facet Designs

作　　者：黎光明 王幸君[1,2] 潘语熙 Li Guangming;Wang Xingjun;Pan Yuxi(School of Psychology,South China Normal University,Guangzhou 510631,China;Center for Studies of Psychological Application,South China Normal University,Guangzhou 510631,China)

机构地区：[1]华南师范大学心理学院,广州510631 [2]华南师范大学心理应用研究中心,广州510631

出　　处：《统计与决策》2022年第3期50-55,共6页Statistics & Decision

基　　金：广东省自然科学基金面上项目(2021A1515012516)。

摘　　要：文章针对正态分布数据,对比Traditional方法、Bootstrap方法和MCMC方法在两侧面交叉设计(p×i×h)和两侧面嵌套设计(p×(i:h))下各个方差分量的估计精度,为实际应用提供参考。使用R软件模拟1000批数据,并在R软件上实现三种方法的方差分量及其变异量估计。结果表明:(1)相较于Traditional方法和MCMC方法,相同条件下,Bootstrap方法估计的方差分量及其变异量结果更为理想;(2)对于两侧面交叉设计和两侧面嵌套设计,在正态分布数据下,建议优先使用Bootstrap方法。Aiming at normal distribution data, this paper compares the estimation accuracy of each variance component of Traditional method, Bootstrap method and MCMC method under two-facet cross design(p × i × h) and two-facet nested design( p ×(i:h)), which provides a reference for practical application. R software is used to simulate 1000 batches of data, and the variance component and variance estimation of the three methods are realized on R software. The results are as follows:(1) Compared with Traditional method and MCMC method, the results of variance component and variation estimated by Bootstrap method are more ideal under the same conditions;(2) The Bootstrap method is recommended to be used in the case of normally distributed data for the two-facet cross design and two-facet nested design.

关 键 词：概化理论 方差分量估计 方差分量变异量估计 BOOTSTRAP方法 MCMC方法 

分 类 号：F224.7[经济管理—国民经济]