Incorrect statistical methods are often used for the analysis of ordinal response data. Such data are frequently summarized into mean scores for comparisons, a fallacious practice because ordinal data are inherently not equidistant. The ubiquitous Pearson chi-square test is invalid because it ignores the ranking of ordinal data. Although some of the non-parametric statistical methods take into account the ordering of ordinal data, these methods do not accommodate statistical adjustment of confounding or assessment of effect modification, two overriding analytic goals in virtually all etiologic inference in biology and medicine. The cumulative logit model is eminently suitable for the analysis of ordinal response data. This multivariate method not only considers the ranked order inherent in ordinal response data, but it also allows adjustment of confounding and assessment of effect modification based on modest sample size. A non-technical account of the cumulative logit model is given and its applications are illustrated by two research examples. The SAS programs for the data analysis of the research examples are available from the author.