Table 2.
Summary results on outlier method performance for measurement and trajectories
| Sensitivity | Specificity | Precision | Kappa | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Min value | Experiment with min value | Max value | Experiment with max value | Min value | Experiment with min value | Max value | Experiment with max value | Min value | Experiment with min value | Max value | Experiment with max value | Min value | Experiment with min value | Max value | Experiment with max value | |
| sBIV | 0.2% | CTX, Type a, 0.5, zMUAC | 99.6% | TK, Type b, 5, zWFL | 98.8% | multiple settings | 99.8% | multiple settings | 2.4% | CTX, Type a, 0.5, zWA | 98.5% | CTX, ALL, 5, zMUAC | 0.002894 | CTX, Type a, 0.5, zMUAC | 0.967156 | TK, Type b, 5, zWFL |
| mBIV | 0.06% | CTX, Type a, 0.5, zWA | 98.4% | TK, Type b, 5, zWFL | 99.7% | multiple settings | 99.9% | multiple settings | 3.5% | CTX, Type a, 0.5, zWA | 99.5% | CTX, ALL, 5, zMUAC | 0.000735 | CTX, Type a, 0.5, zWA | 0.96498 | TK, Type b, 5, zWFL |
| SMOM | 5.6% | CTX, Type c, 0.5, zWA | 99.6% | TK, Type b, 5, zWFL | 94.9% | CTX, Type a, 0.5, zMUAC | 99.9% | CTX, ALL, 5, zMUAC | 2.4% | CTX, Type c, 0.5, zWA | 99.4% | CTX, ALL, 5, zMUAC | 0.006147 | CTX, Type c, 0.5, zWA | 0.930517 | TK, Type b, 5, zWFL |
| MMOM | 5.7% | CTX, Type c, 0.5, zWA | 99.0% | CTX, Type a, 5, zWA | 95.2% | CTX, Type c, 0.5, zWFL | 99.8% | CTX, ALL, 5, zMUAC | 2.4% | CTX, Type c, 0.5, zWA | 98.7% | CTX, ALL, 5, zMUAC | 0.005508 | CTX, Type c, 0.5, zWA | 0.717394 | CTX, Type a, 4, zMUAC |
| COT | 25.8% | TK, Type a, 0.5, zWFL | 99.8% | CTX, Type a, 5, zWA | 73.8% | CTX, Type b, 0.5, zWA | 98.9% | CTX, ALL, 5, zMUAC | 14.9% | CTX, Type c, 0.5, zWA | 99.1% | CTX, ALL, 5, zMUAC | -0.00023 | CTX, Type c, 0.5, zWA | 0.906329 | CTX, Type a, 5, zWA |
| MMOT | 2.3% | TK, ALL, 4, zWFL | 24.9% | CTX, Type a, 5, zWA | 93.9% | TK, Type b, 0.5, zWFL | 99.9% | TK, ALL, 5, zWFL | 13.1% | CTX, Type b, 0.5, zWFL | 99.8% | TK, ALL, 5, zWFL | -0.00769 | CTX, Type b, 0.5, zWA | 0.3212.91 | CTX, Type a, 5, zWA |
Abbreviations: TARGet Kids! The applied research group for kids, zWFL Weight-for-length z-scores, zWA weight-for-age z-scores, CTX The co-trimoxazole prophylaxis trial, zMUAC mid-upper arm circumference-for-age z-scores, SD Standard deviation, mBIV Modified method for biologically implausible values detection, sBIV Static WHO cut-off values for biologically implausible values detection method, MMOM Multi-model outlier measurement detection method, SMOM Single-model outlier measurement detection method, COT Clustering-based outlier trajectory detection method, MMOT Multi-model outlier trajectory detection method
The "setting" of minimum or maximum values corresponds to a tuple specifying the dataset (CTX or TARGet Kids!), the type of the error, the intensity of the error in SD, and the anthropometric measure