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

Table 1 Key characteristics of the employed outlier detection methods

Method	Key characteristics
Method	Type of detection method	Types of outliers	Input parameters	Advantages
Static BIV (sBIV)	Standardized	Measurements	Fixed cut-offs	Simple
Modified BIV (mBIV)	Empirical	Measurements	Fixed cut-offs	Time consensus, simple
Single-model outlier measurement detection (SMOM)	Statistical based	Measurements	Semi-dynamic^a, based on the dataset	Population adjusted
Multi-model outlier measurement detection (MMOM)	Statistical and clustering based	Measurements	Semi-dynamic, based on the dataset	Group-adjusted
Clustering-based outlier trajectory (COT)	Clustering based	Trajectory	Dynamic, based on data size	Population adjusted
Multi-model outlier trajectory (MMOT)	Statistical and clustering based	Trajectory	Semi-dynamic, based on the dataset	Group-adjusted

^aCombination of fixed thresholds (i.e. WHO) and dynamic thresholds or values derived from the dataset (including averages, number of clusters and so on)

ISSN: 1471-2288