May W L, Johnson W D
Department of Biometry and Genetics, Louisiana State University Medical Center, New Orleans 70112-1393, USA.
J Biopharm Stat. 1996 May;6(2):139-53. doi: 10.1080/10543409608835129.
This paper addresses the problem of analyzing multivariate response with the variates having different distributions. We use the generalized estimating equations proposed by Prentice and Zhao (1) to estimate mean and covariance parameters. Wald statistics are used to test hypotheses about the parameters. Data from a two-group study of competing programs to increase memory recall in geriatric patients are analyzed. We are interested in determining whether there is a simultaneous increase in test scores with one variate being continuous and another being dichotomous. In a second example, comparisons of before-and-after treatment serum cholesterol levels and number of snacks eaten per day are made in an experiment to determine efficacy of a program aimed at simultaneously reducing both measurements.
本文探讨了分析具有不同分布的多个响应变量的问题。我们使用普伦蒂斯和赵（1）提出的广义估计方程来估计均值和协方差参数。使用Wald统计量来检验关于参数的假设。分析了一项针对老年患者提高记忆回忆能力的两组竞争性方案研究的数据。我们感兴趣的是确定在一个变量为连续型而另一个变量为二分变量的情况下，测试分数是否会同时提高。在第二个例子中，在一项旨在同时降低这两个测量值的方案疗效的实验中，对治疗前后的血清胆固醇水平和每天吃零食的数量进行了比较。
