AN ALTERNATIVE TO THE COCHRAN-(Q) STATISTIC FOR ANALYSIS OF HETEROGENEITY IN META-ANALYSIS OF DIAGNOSTIC TESTS BASED ON HJ BIPLOT

Abstract

he possible heterogeneity among individual studies constituting a meta-analysis is traditionally evaluated using Cochran Q and Higgins I 2 statistics. However, both indices have deficiencies: The Q statistic detects heterogeneity but does not allow its quantification, whereas the I 2 index allows for quantification of heterogeneity but does not indicate which studies are responsible for it. This problem is solved by additionally using the HJ biplot of the matrix containing the information about true positives, true negatives, false positives, and false negatives for each study. This means that the information contained in such a tetrachoric table contains the joint frequency distribution of the true classification of the disease and that provided by the diagnostic test