Fig. 3.

PLS regression model for fat prediction by HSI in single (a,b) shelled and (c) in-shell cocoa beans. a) Regression coefficients for fat expressed on “as is” or dry matter basis for the shelled beans. b) Latent Variable plot to express the calibration, cross-validation and external prediction error (RMSE). Models use MSC spectral pre-treatment. c) Regression coefficient plot for in-shell beans. Numbers indicate the wavelength in nm. The arrow indicates the selected optimal number of Latent Variables.