McNemar test doesn't produce a test statistic
I'm running a McNemar test in CROSSTABS, and all I get is a probability. I don't see a chi-square value. Why not?
Resolving the problem
CROSSTABS computes the exact binomial probability rather than computing a significance based on an asymptotic chi-square approximation of a statistic to a chi-square distribution. There is thus no test statistic involved.
If you have a reason to absolutely need the chi-square approach, the NPTESTS procedure (Analyze>Nonparametric Tests>Related Samples in the menus) always produces the asypmtotic chi-square statistic and associated significance. If the number of off-diagonal elements of the 2x2 table is 25 or fewer, the results in the Hypothesis Test Summary table shown initially in the output are based on the exact binomial-based probability. However, if you double-click the object to open the Model Viewer, the table for the McNemar test in the Related Samples Test view on the right shows the (continuity-corrected) chi-square test statistic and associated asymptotic significance.
Translate this page: