Fig. 3

Example of training data. Classified subset data \({\varvec{X}}^{Imposed}\) (top), where rows are metabolites, columns are samples, and colors indicate different metabolites used to create the training data (bottom). Columns in the bottom table indicate the metabolite abundance, different features at the metabolite level (max, min, etc.), and quantile level for the abundance level. The class-label of interest is Target which is the missing data type: MCAR, MNAR, or O (signifies the Other nuisance class for non-missing entries). Colors indicate the original metabolites in \({\varvec{X}}^{Imposed}\) (top)