Fig. 3
From: Fingerprinting cities: differentiating subway microbiome functionality
Distribution of prediction scores from the city predictors trained on the selected 20 ECs. a AKL (Auckland); b HAM (Hamilton); c NYC (New York City); d OFA (Ofa); e PXO (Porto); f SAC (Sacramento); g SCL (Santiago); h TOK (Tokyo). Positive (P) and negative (N) score distributions for raw-select models were less obvious to their resampled model (balance-select) versions (ReP and ReN)