An Expanded McNemar Approach for Evaluation of Longitudinal Data

The standard chi-square test, developed by Pearson (1900) determines whether significant differences exist between frequencies of data categories, each containing unique subjects. In other words, each subject provides data for only one category. This stipulation prevents analysis of longitudinal data pertaining to subjects’ categorization at more than one point in time. The McNemar test (1947) produces a chi-square value that, in contrast to the Pearson chi-square value, accounts for dichotomous data from a single sample measured twice. Although extremely beneficial for data analysis, the McNemar test cannot be used for situations involving more than two measurements of the sample. This article, therefore, establishes a formula to compare dichotomous data from three or more trials involving the same subjects. An example demonstrates the formula’s applicability in a three-trial situation. Further expansion of the formula used for this example, as described in subsequent text, makes it possible to evaluate differences in repeated-measures frequencies for any number of trials.