Gender discrimination may be worse than you think: testing ordinal interactions in power research

The authors reanalyze the data of a study by S. M. Elias and R. J. Loomis (2004), which aimed to determine how an instructor's gender may influence his or her ability to gain student compliance. S. M. Elias and R. J. Loomis observed few significant gender effects using traditional multivariate analyses of variance. The authors reanalyze this data using the more appropriate statistical techniques for detecting ordinal interactions recommended by M. J. Strube and P. Bobko (1989) and S. M. Elias (2004). An ordinal interaction occurs when 1 cell of a 2 x 2 design is responsible for a significant interaction (e.g., female instructors suffering only when rated by male students). Reanalysis of the data resulted in more robust findings.