Click to collapse section Community Description ▼ Click to expand section Community Description ▶ Community Description
SPSS v 20 HELP says that McNemar's test is computed using the binomial distribution if the total N <25, but uses the chi-squared approximation if N >= 25. HOWEVER no matter what the N, the output always fails to show a value for chi-squared, and the p-value is always footnoted to say that the binomial distribution was used to obtain the p-value. By my computations it is correct to say that inspite of their claims, IBM uses the binomial distribution in all cases, regardless of N, to calculate the p-value for McNemar's test. They should certain correct the program CROSS-TABS so that the output matches the HELP claims ... or, much better, they should leave the HELP claims as is and correct the program to use chi-squared when N is large.