Fig. 1

The proposed Bayesian models have a much better 95 % CI coverage than the standard logistic regression model. 95 % confidence/credible interval (CI) coverage for four different methods are shown for different scenarios of sensitivity (SN) and specificity (SP) [SN = 0.6 and SP = 0.9 (upper left panel); SN = 0.9 and SP = 0.9 (upper right panel); SN = 0.6 and SP = 0.98 (lower left panel); SN = 0.9 and SP = 0.98 (lower right panel)]. These results are based on 100 simulated datasets, with 2000 individuals in each dataset. Results closer to 0.95 (blue horizontal dashed lines) indicate better performance