析因设计资料的单独效应分析  

Analysis of Simple Effect on Factorial Design

作　　者：谭旭辉[1] 钱俊[1] 陈平雁[1] 

机构地区：[1]南方医科大学公共卫生与热带医学学院生物统计学系,510515

出　　处：《中国卫生统计》2007年第4期378-379,382,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金(30571621);广东省科技计划项目(2005B33701005)

摘　　要：目的探讨析因设计资料单独效应的分析方法。方法分析单独效应的两种方法分别为基于析因分析的整体标准误和基于单因素分析的局部标准误方法,前者使用SPSS软件中一般线性模型下Lmatrix子句实现。用模拟方法分别比较两种方法的I型错误率和检验效能。结果使用两种不同标准误,统计分析结果存在较大差异。模拟结果显示,两种方法的I型错误率非常接近,基于析因方差分析标准误的检验效能较高。结论基于析因分析整体标准误较基于单因素局部标准误分析单独效应更为合理。Objective To explore the methods of analysis of simple effect on factorial design. Methods Two methods, one was based on the mean square of error term coming from factorial analysis （SE1） and another from one-way ANOVA （SE2）, were employed to analyze simple effect on factorial design. The former was conducted under Lmatrix subeom- mand in GLM. The simulating method was used to compare the type I error and testing power. Results Considerable differences were manifested between the results from two methods based on the real example. The simulation results showed that the type Ⅰ errors were nearly equal to each other, and the method based on SE1 had higher testing power. Conclusion The method based on SE1 for simple effect analysis under factorial design was more reasonable than one based on SE2.

关 键 词：析因设计 交互作用 单独效应 一般线性模型 

分 类 号：R195[医药卫生—卫生统计学]