交叉设计资料的混合效应模型分析

Mixed Effects Model for a General Cross-over Design

目的放宽交叉设计方差分析对残留效应假定和探讨交叉设计混合效应分析模型。方法建立三个模型以适应残留效应的不同假定,即残留效应为零、相等或不等,通过构造对数似然函数以及利用Fisher记分迭代算法可求得β和θ的极大似然估计与限制极大似然估计。结果用SAS程序实现了交叉设计混合效应模型分析,得到了有关参数的估计值和直接处理效应的比较结果,实例分析表明,混合效应模型能够提供更多的有效信息。结论混合效应模型较一般方差分析有更强的适应性,可完善和丰富交叉设计资料的分析方法。

Objective To relax assumption of carryover effects in analysis of variance for cross - over designs and explore mixed effects model. Methods Based on the various assumptions about the carryover effects, i.e., whether they are null, equal or unequal, three statistical models are presented. The log - likelihood function is formulated, and then,maximum likelihood （ML） estimation and restricted maximum likelihood （REML） estimation ofandare found by using Fisher scoring iterative algorithm. Results The estimators for model parameters and comparisons of direct treatment effects are got by submitting SAS programs in mixed effects model analysis, and more useful information is provided by this model,along with a medical example. Conclusion The adaptation of mixed effects model is more flexible than general ANOVA, mixed effects model would provide efficient clues and techniques for analyzing the data in cross-over designs.