Nonparametric Statistics Notes - Cochran Q Test

The Cochran Q test is an extension of McNemar's Chi-square test for changes (see the 2 x 2 Tables option accessed from the Nonparametric Statistics Startup Panel - Quick tab) in frequencies or proportions to k (more than two) dependent samples. Specifically, it tests whether several matched frequencies or proportions differ significantly among themselves. Select Cochran Q test from the Nonparametric Statistics Startup Panel - Quick tab to display the Cochran Q Test dialog box, in which you select a variable list and codes that identify the two categories or levels of the dichotomous measure. The test assumes that the variables are coded as 1's and 0's, and the codes you have specified will cause the variables to be transformed accordingly (only for this analysis, the data file itself will not be changed).

Assumptions and interpretation. The Cochran Q test only requires a nominal scale, or that the data have been artificially dichotomized. A typical example where the Q test is useful is when you want to compare the difficulty of dichotomous questionnaire items that can either be answered right or wrong. Here, each variable in the data file would represent one item, and contain 0's (wrong) and 1's (right). If the Q test is significant, then we conclude that the items are of different difficulty since different items were answered correctly by more or fewer respondents.

See also, Cochran Q Test - Quick tab for further details.