Fig. 6

Computation time comparison. Panel a shows boxplots of the computation time for the original estimator for all 9 experiments grouped by their sample sizes and color coded according to the proportion of negative and positive cases. Panel b shows the analogous results for the proposed estimator. Note the different scales in the y-axis of panels a and b. The boxplots report the computation time across the same 1,000 replications of each one of the 9 experiments presented in Fig. 5 (and described in Table 1)