Fig. 3.

PCA of all 107 MSC stool µ_a, µ_s’ and g spectra (a,b and c). A probabilistic cluster analysis was carried out on the scores of the first two principal components (PC), delivering two clusters for each property. In case of the µ_a spectra, these clusters coincide for 90% of all stool probes with the classification of stool into meconium (blue) and transitional stool (red). Following the development of µ_a from meconium to transitional stool, the clustering predicts a correct course for 17 out of 25 children (d).