Studies in the Design and Analysis of Dental Experiments

5. Non-parametric Tests (Correlated Samples with Dichotomous Scales)

¹ School of Public Health and Administrative Medicine of the Faculty of Medicine, Columbia University, New York, New York

Two relatively new tests for data classified according to a dichotomous scale have been described. These were tests for the case of correlated samples: the Quinn McNemar test for two samples and the Cochran Q test for more than two samples. These tests were illustrated with data on topical anesthesia of oral mucous membranes in which each individual served as his own control. An analysis-of-variance approach has been described that facilitates the calculation of appropriate significance tests not only for the case of correlated samples but also for the case of independent samples. This technique also affords measures of association between the responses of the matched samples.