合理进行均值比较——具有组间异质性的增长重复测量模型

Reasonably carry out mean value comparison:repeated measurement model of the growth with inter-group heterogeneity

本文目的是介绍与具有组间异质性的增长重复测量模型有关的5个基本概念、计算方法、一个临床试验实例及其SAS实现。基本概念包括重复试验与重复测量、方差协方差结构、均衡性与异质性、先验分布与后验分布、混合模型与多水平模型。计算方法涉及抽样方法和先验分布。临床试验实例涉及具有异质性的两组增长重复测量数据。本文呈现了采用SAS处理实例中增长重复测量资料的全过程,包括呈现两组儿童口腔内一特定指标(简称距离)随年龄变化的增长图、构建两组儿童线性重复测量模型、构建一个既考虑到异质性又能估计性别效应的模型、直接获得与性别有关的截距和斜率的后验分布,以及基于后验估计值直接推断概率。对输出结果作出了解释,并讨论了改用多水平模型所产生的效果。

The purpose of this article was to introduce 5 basic concepts,calculation methods,a clinical trial example and its SAS implementation related to the repeated measurement model of growth with intergroup heterogeneity.The basic concepts included the repeated experiments versus repeated measurements,variance-covariance structures,equilibrium versus heterogeneity,prior versus posterior distributions,mixed versus multilevel models.The computational methods involved the sampling methods and the prior distributions.The clinical trial example involved two groups of repeated measurement of growth with heterogeneity.This paper gave the whole process of using SAS software to analyze the growth repeated measurement data in the example,including presenting the growth graph of a specific index(short for distance)in the mouth of the two groups of children with age,constructing the linear repeated measurement model of the two groups of children,constructing a model that took into account heterogeneity while estimating gender effects,directly obtaining posterior distributions for gender-dependent intercepts and slopes,and directly inferring probabilities based on the posterior estimates.The output results were explained and the effects of switching to a multilevel model were discussed.