Abstract

Several arguments are presented supporting the more extensive use of parametric statistics. These arguments are concerned with (1) the insensitivity of ordinal and other nonparametric techniques, (2) the small error that results from assigning numbers to ordinal data and then treating the categories as if they conform to an interval scale, (3) tests of statistical robustness, and (4) the power-efficiency of tests. The utility of assigning numbers or scoring systems to ordinal data and then using intervally-based statistics is supported in several ways. In addition, it is demonstrated that the linear scoring system results in a small amount of error no matter what the “true” scoring system may be. However, the assignment of any scoring system must be consistent with the monotonic function.

You do not currently have access to this article.