Comparison of Predictive Values from Two Diagnostic Tests in Large Samples

SUMMARY Within the field of diagnostic tests, the positive predictive value is the probability of being diseased given that the diagnostic test is positive. Two diagnostic tests are applied to each subject in a study and in this report we look at statistical hypothesis tests for large samples to compare the positive predictive values of the two diagnostic tests. We propose a likelihood ratio test and a restricted difference test, and we perform simulation experiments to compare these tests with existing tests. For comparing the negative predictive values of the diagnostic tests, i.e. the probability of not being diseased given that the test is negative, we propose negative predictive versions of the same tests. The simulation experiments show that the restricted difference test performs well in terms of test size.