Classification trees as proxies

In data analysis, when data are unattainable, it is common to select a closely related attribute as a proxy. But sometimes substitution of one attribute for another is not sufficient to satisfy the needs of the analysis. In these cases, a classification model based on one dataset can be investigated as a possible proxy for another closely related domain's dataset. If the model's structure is sufficient to classify data from the related domain, the model can be used as a proxy tree. Such a proxy tree also provides an alternative characterization of the related domain. Just as important, if the original model does not successfully classify the related domain data the domains are not as closely related as believed. This paper presents a methodology for evaluating datasets as proxies along with three cases that demonstrate the methodology and the three types of results.