Table 3. Posterior probability of event or non-event given the number of positive ratings.
Total Number of Adverse Event ratings out of 30 raters | Probability that case truly is an Adverse Event |
0 (0%) | 0.0000 |
1 (3.3%) | 0.0000 |
2 (6.6%) | 0.0000 |
3 (9.9%) | 0.0000 |
4 (13.2%) | 0.0000 |
5 (16.5%) | 0.0000 |
6 (20.0%) | 0.0000 |
7 (23.3%) | 0.0000 |
8 (26.7%) | 0.0003 |
9 (30.0%) | 0.0026 |
10 (33.3%) | 0.0219 |
11 (36.7%) | 0.1612 |
12 (40.0%) | 0.6225 |
13 (43.3%) | 0.9340 |
14 (46.7%) | 0.9918 |
15 (50.0%) | 0.9990 |
16 (53.3%) | 0.9999 |
17 (56.7%) | 1.0000 |
18 (60.0%) | 1.0000 |
19 (63.3%) | 1.0000 |
20 (66.7%) | 1.0000 |
21 (70.0%) | 1.0000 |
22 (73.3%) | 1.0000 |
23 (76.7%) | 1.0000 |
24 (80.0%) | 1.0000 |
25 (83.3%) | 1.0000 |
26 (86.7%) | 1.0000 |
27 (90.0%) | 1.0000 |
28 (93.3%) | 1.0000 |
29 (96.7%) | 1.0000 |
30 (100%) | 1.0000 |
These estimates are from the latent class model that assumed a common sensitivity and specificity for all reviewers.