Common Errors in Statistics (and How to Avoid Them), 4th Edition

ORDINAL DATA

Ordinal data include measurements but also include observations that, while ordered, cannot be usefully added and subtracted as measurements can. Observations recorded on the familiar Likert scale of 1-Disliked Intensely to 9-Liked Very Much, with 5 representing Indifference, are an example of ordinal but nonmetric data. One cannot assume that the difference between Disliked Intensely (1) and Disliked (3) is the same as between Disliked (3) and Indifferent (5). Thus, an arithmetic average or a variance would not be at all meaningful.

One can report such results in tabular form, in bar charts, or by providing key percentiles such as the minimum, median, and maximum. Contrary to other published recommendations (e.g., Porter, 1999), the Pearson correlation coefficient can be used with ordinal data (Good, 2009).