Min and Max Scorings for Two-Sample Ordinal Data

Abstract

Abstract To analyze two-sample ordinal data, one must often assign some increasing numerical scores to the ordinal categories. The choice of appropriate scores in these types of analyses is often problematic. This article presents a new approach for reporting the results of such analyses. Using techniques of order-restricted inference, we obtain the minimum and maximum of standard two-sample test statistics over all possible assignments of increasing scores. If the range of the min and max values does not include the critical value for the test statistics, then we can immediately conclude that the result of the analysis remains the same no matter what choice of increasing scores is used. On the other hand, if the range includes a critical value, the choice of scores used in the analysis must be carefully justified. Numerous examples are given to clarify our approach.