McGee V E
Multivariate Behav Res. 1968 Apr 1;3(2):233-48. doi: 10.1207/s15327906mbr0302_7.
Given N sets of similarity measures between pairs of stimulus objects a nonmetric multidimensional scaling procedure (known as CEMD) is proposed which allows one of four kinds of analysis to be performed on the data. The four types of solution are defined in terms of two binary decisions: (a) whether only one configuration is allowed for all N sets or whether each set is allowed to have its own configuration; and (b) whether the same mono- tonic transformation is applied to all N sets of input data or whether each set is allowed its own monotonic transformation. Where separate configurations are allowed the computer program seeks to keep these solutions as similar as possible and provides an index of the similarity of the configurations.
给定刺激对象对之间的N组相似性度量，提出了一种非度量多维缩放程序（称为CEMD），该程序允许对数据进行四种分析之一。这四种解决方案类型是根据两个二元决策定义的：（a）对于所有N组，是只允许一种配置，还是每组都允许有自己的配置；（b）对于所有N组输入数据，是应用相同的单调变换，还是每组都允许有自己的单调变换。在允许有单独配置的情况下，计算机程序会试图使这些解决方案尽可能相似，并提供配置相似性的指标。
