Fig 5. Principle component analysis (PCA) of a database containing 324 sample vectors in a 36-dimensional space.

Each vector was transformed via PCA into three mutually orthogonal principle components, which was plotted according to different categories: (a) bronchial constriction level, (b) inhalation flow rate, (c) particle size, and (d) upper airway variation. Varying degrees of data clustering exist among the four categories, with the particle sizes showing nearly no separation.