Table 4.

Multiple Linear Regression (MLR) models for moisture and fat content (% “as is” basis) in green coffee beans obtained by selecting the most important variables using log (1/R) values with no spectral pre-treatment.

Parameter	Pre-treatment	Spectral variables nr.	Calibration		Cross-validation		Validation Offset
			R2	RMSE	R2	RMSECV
Moisture	log (1/R)	6a	0.615	0.459	0.596	0.471	0.61
		42	0.887	0.242	0.842	0.286	0.29
Fat		6a	0.693	1.464	0.676	1.503	4.57
		6b	0.841	1.053	0.834	1.078	2.36
		22	0.860	0.988	0.832	1.081	2.17

^aVariable selection was performed based on the β-coefficients of the PLS model (filter method).

^bModel built by PLS regression (LV = 6).