New approaches and technical considerations in detecting outlier measurements and trajectories in longitudinal children growth data

Table 4 Key findings

1. Clustering-based outlier trajectory detection (COT) is a reliable method for outlier trajectory detection.
2. Combined detection methods for outlier measurements are preferred.
3. Some methods achieved >80% sensitivity for errors above 3 standard deviations
4. Model-based methods are reliable for errors of lower intensity.
5. Higher density favours outlier trajectory detection and model-based methods, but not time-sensitive methods.
6. Clustering and pattern analyses can be considerably affected by the presence of outliers.

ISSN: 1471-2288