Statistics USING THE SAS ® SYSTEM TO PERFORM MCNEMAR ' S TEST AND CALCULATE THE KAPPA STATISTIC FOR MATCHED PAIRS OF DATA

Dichotomous responses from matched pairs of data at different timepoints are often displayed in 2 by 2 tables, and McNemar's test is performed to test the hypothesis that the proportion of subjects with a certain response is equivalent at both timepoints. A kappa statistic can determine the level of agreement in these responses. The SAS procedure FREQ can perform a large sample approximation to McNemar's test when the number of discordant matched pairs is large (n > 20). The CMHI option in FREQ is used since McNemar's test is a special case of the Cochran-ManteI-Haensze1 statistic. In SAS Version 6.11, McNemar's test can also be performed using the new option AGREE in FREQ. If the number of discordant matched pairs is small (n ~ 20), the SAS procedure UNIVARIATE will perform an exact McNemar's test. An example is presented performing McNemar's test using the procedure UNIVARIATE. McNemar's test is performed to determine whether the proportion of patients considered to be eradicated of H.py/ori infection is equivalent at two different timepoints. An algorithm and SAS code for calculating a kappa statistic and 95% confidence interval were developed and will be presented.