Figure 22 presents histograms of absolute differences in estimated sensitivity and specificity in univariate vs. bivariate random effects inverse variance meta-analyses (both using a normal approximation for within-study variability). The plots for sensitivity (left panel) and specificity (right panel) compare bivariate vs. univariate random effects inverse variance meta-analyses (Der-Simonian-Laird vs. multivariate DerSimonian-Laird; both using a normal approximation for within-study variability). Both graphs indicate that absolute differences are generally small (in a single meta-analysis the difference was beyond 0.05, and only for specificity).

Figure 22Histograms of differences in estimated summary sensitivity and specificity in univariate versus bivariate random effects inverse variance meta-analyses (both using a normal approximation for within-study variability)

Note: Histograms of differences in estimated summary sensitivity (left panel) and specificity (right panel) comparing bivariate versus univariate random effects inverse variance meta-analyses (Der-Simonian-Laird vs. multivariate DerSimonian-Laird; both using a normal approximation for within-study variability).

DL = DerSimonian-Laird; mult. DL = multivariate DerSimonian-Laird.

From: Results

Cover of An Empirical Assessment of Bivariate Methods for Meta-Analysis of Test Accuracy
An Empirical Assessment of Bivariate Methods for Meta-Analysis of Test Accuracy [Internet].
Dahabreh IJ, Trikalinos TA, Lau J, et al.

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.