ABSTRACT
Ordinal data come in several forms. One form consists of ranking the cases in a sample from high to low on some attribute, such as creativity, motivation, and so forth. Another form is ordered categories, where cases are classified into three or more groups that have some natural ordering, such as high, medium, and low. A third form consists of scores from a rubric, such as the one shown in Table 6.1. In this rubric, the quality of the arguments in students’ essays was rated on a 7-point scale. While it is tempting to analyze these scores as metric, that assumes that the distances between the adjacent points on the scale are roughly equal on some underlying continuous variable. That assumption might be reasonable in some circumstances, but when it is not, treating the scale as metric can introduce additional measurement error into the scores, thereby reducing statistical power. If in doubt, it is safest to treat rubric scores as ordinal.
