To the Eitor: Moses et al. recommend analyzing data with the Mann–Whitney U test or a t-test, to take account of ordinal measurement (Aug. 16 issue).1 Their motivation is that "[i]nappropriate analysis of such data (for example, reducing them to a two-point scale indicating presence or absence) can sacrifice much of the information in the data." Their recommendation, however, encourages authors to employ a statistical practice — testing for statistical significance — that discards nearly all the information in a set of data.Moses et al. signal clearly their preoccupation with the presence or absence of statistical significance: "If the . . .