Department of Statistics, Uppsala University, Uppsala, Sweden.
Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Stat Med. 2018 Sep 20;37(21):3078-3090. doi: 10.1002/sim.7690. Epub 2018 Jun 11.
In the analysis of ordered categorical data, the categories are often assigned a set of subjectively chosen order-restricted scores. To overcome the arbitrariness involved in the assignment of the scores, several score-independent tests have been proposed. However, these methods are limited to 2 × K contingency tables, where K is the number of ordered categories. We present an efficiency robust score-independent test that is applicable to more general situations. The test is embedded into a flexible framework for conditional inference and provides a natural generalization of many familiar tests involving ordered categorical data, such as the generalized Cochran-Mantel-Haenszel test for singly or doubly ordered contingency tables, the Page test for randomized block designs and the Tarone-Ware trend test for survival data. The proposed method is illustrated by several numerical examples.
在有序分类数据的分析中，类别通常被赋予一组主观选择的受限制的分数。为了克服分数赋值中的任意性，已经提出了几种不依赖于分数的检验方法。然而，这些方法仅限于 2×K 列联表，其中 K 是有序类别的数量。我们提出了一种适用于更一般情况的有效稳健的不依赖于分数的检验方法。该检验被嵌入到一个灵活的条件推断框架中，并为许多涉及有序分类数据的熟悉检验方法提供了自然的推广，例如用于单或双有序列联表的广义 Cochran-Mantel-Haenszel 检验、用于随机区组设计的 Page 检验以及用于生存数据的 Tarone-Ware 趋势检验。所提出的方法通过几个数值例子进行了说明。
