Table 2. Summary of performance metrics and significance measures for 100 RF models of each of 14 combinations of vehicle pairs, interpolation/extrapolation and AUC significance. Explanation of interpolation/extrapolation, AUC, N, equivocal and contradictory is given in Table 1 .
| Vehicle pair | Interpolation/extrapolation | AUC/% | N (less toxic in first vehicle : second vehicle) | Equivocal | Contradictory | Mean balanced accuracy/% | Mean probability of real model being better than random/% (SD) | Mean overlap between real and random models (SD) |
| Saline vs. Water | Extrapolate high | 30 | 50 (29 : 21) | 54 | 12 | 73 | 91 (20) | 9 (16) |
| Saline vs. saline with Tween-80 | Extrapolate high-low | 40 | 119 (62 : 57) | 53 | 6 | 63 | 85 (23) | 16 (18) |
| Saline vs. MC | Interpolation only | 30 | 60 (39 : 21) | 19 | 3 | 69 | 81 (24) | 17 (17) |
| Saline vs. HPC | Interpolation only | 40 | 68 (36 : 32) | 32 | 0 | 70 | 81 (28) | 13 (16) |
| Saline vs. CMC | Interpolation only | 60 | 66 (43 : 23) | 55 | 2 | 71 | 94 (17) | 8 (14) |
| Saline vs. CMC | Interpolation only | 40 | 108 (68 : 40) | 87 | 9 | 63 | 84 (27) | 15 (16) |
| Saline vs. CMC | Interpolation only | 30 | 144 (95 : 49) | 106 | 18 | 62 | 90 (21) | 12 (16) |
| Saline vs. CMC | Extrapolate high | 60 | 92 (58 : 34) | 70 | 6 | 79 | 100 (1) | 0 (0) |
| Saline vs. CMC | Extrapolate high | 40 | 130 (78 : 52) | 99 | 15 | 63 | 91 (21) | 13 (20) |
| Saline vs. CMC | Extrapolate high | 30 | 164 (103 : 61) | 112 | 26 | 61 | 89 (22) | 13 (17) |
| Saline vs. CMC | Extrapolate high-low | 60 | 79 (49 : 30) | 76 | 4 | 73 | 98 (11) | 3 (7) |
| Saline vs. CMC | Extrapolate high-low | 40 | 123 (69 : 54) | 100 | 14 | 66 | 94 (14) | 10 (15) |
| HPC vs. saline with Tween-80 | Interpolation only | 40 | 62 (26 : 36) | 32 | 5 | 71 | 88 (23) | 7 (9) |
| HPC vs. saline with Tween-80 | Extrapolate high-low | 30 | 107 (56 : 51) | 45 | 7 | 66 | 88 (23) | 7 (12) |