Are ordinal models useful for classification? a revised analysis

Publisher Website
Google Scholar
Abstract
This paper describes a simulation study comparing the performances of various ordered and unordered models, in terms of their error rates, to those given by Campbell et al. (1991), but only applying those models which are appropriate to the particular data set generated. By using the exact error rates and only fitting appropriate models, we reach the opposite conclusions to those of Campbell et al. (1991). Furthermore, by generating data sets using other underlying models, we show that models which reflect the underlying data structure perform better than those which do not.