Table 2.
Frequency of the MCHS-P, MCS-P, 50% MWS, and Gini coefficient obtaining the optimal result* in 20 multi-clusters scenarios and in 30 single-cluster scenarios for the five classic performance measures.
| Scenario type | In 20 multi-cluster scenarios | In 30 single-cluster scenarios | ||||||
|---|---|---|---|---|---|---|---|---|
| Measures | MCHS-P | MCS-P | Default | Gini | MCHS-P | MCS-P | Default | Gini |
| Sensitivity | 19 | 0 | 1 | 0 | 26 | 28 | 4 | 4 |
| Specificity | 0 | 20 | 0 | 0 | 26 | 30 | 26 | 26 |
| PPV | 0 | 20 | 0 | 0 | 2 | 5 | 25 | 26 |
| YDI | 20 | 0 | 1 | 1 | 22 | 28 | 2 | 2 |
| Misclassification | 20 | 0 | 4 | 6 | 26 | 30 | 22 | 22 |
*The optimal result is defined as the range that varies within 0.001 around the real optimal value of classic performance measure, such as the maximal sensitivity, specificity, PPV, YDI and the minimal misclassification. Therefore, in a given scenario, it is possible for more than one method to obtain the optimal result.