Classification of Coffee Beans by GC-C-IRMS, GC-MS, and ¹H-NMR

Figure 3

(a) and (b) show the behavior of as a function of the number of latent vectors (LVs) for 100 models sampled randomly for GC-MS (7 predictors, (c)) and NMR (1610 predictors, (b)). The thick red curve represents the average of all models and its turning point is used to determine the best number of components for which the distributions are shown on (c). The green curve represents the distribution of for 2 LVs (GC-MS). The blue and black curves are for NMR using 2 and 8 LVs, respectively, and using either all the predictors (black) or only the best 8 predictors (blue), selected according to their VIP.

(a)

(b)

(c)