Johnson W D, Koch G G
Health Serv Res. 1970 Winter;5(4):358-69.
A procedure is described for fitting linear models to categorical data obtained by stratified random sampling. The method makes it possible to compute estimates of variance and to test certain types of hypotheses not testable for such data by the usual chisquare methods. Details of the method are illustrated in application to data from a survey of hospitals with regard to implementation of and satisfaction with a service program.
本文描述了一种将线性模型拟合到通过分层随机抽样获得的分类数据的方法。该方法能够计算方差估计值,并检验某些类型的假设,而这些假设用常用的卡方方法对此类数据是无法检验的。通过将该方法应用于一项关于医院对一项服务项目的实施情况和满意度的调查数据,对该方法的细节进行了说明。