. 2020 Aug 28;9(9):1193. doi: 10.3390/foods9091193

Table 4.

Partial least squares (PLS) regression analysis for the prediction of well-modelled wine sensory variables from wine chemical data, excluding amino acid analysis ^†.

Sensory Attribute	Data Selected	PLS Model Parameters ^‡
Sensory Attribute	Data Selected	Factor No	R²_cal	R²_val	RMSE_cal	RMSE_val
Opacity	all variables	1	0.82	0.58	0.34	0.58
	sig var ^§	3	0.96	0.58	0.15	0.58
Brown colour	all variables	1	0.92	0.86	0.10	0.14
	sig var	3	0.98	0.94	0.05	0.09
Red fruit aroma	all variables	1	0.61	0.49	0.16	0.21
	sig var	3	0.57	0.38	0.17	0.23
Dark fruit aroma	all variables	1	0.72	0.50	0.20	0.30
	sig var	3	0.96	0.58	0.07	0.28
Dried fruit aroma	all variables	1	0.83	0.73	0.13	0.18
	sig var	3	0.87	0.73	0.13	0.20
Spice aroma	all variables	1	-	-	-	-
	sig var	3	0.79	0.64	0.10	0.14
Earthy aroma	all variables	1	0.79	0.67	0.10	0.13
	sig var	3	0.78	0.62	0.09	0.14
Pepper aroma	all variables	1	0.70	0.53	0.07	0.09
	sig var	3	0.70	0.36	0.07	0.11
Pungent aroma	all variables	1	0.79	0.65	0.11	0.15
	sig var	3	0.87	0.62	0.08	0.16
Sweetness	all variables	1	0.81	0.71	0.10	0.14
	sig var	3	0.96	0.85	0.05	0.10
Viscosity	all variables	1	0.69	0.43	0.16	0.24
	sig var	3	0.89	0.35	0.09	0.25
Dark fruit flavour	all variables	1	0.73	0.43	0.23	0.38
	sig var	3	0.90	0.40	0.14	0.39
Dried fruit flavour	all variables	1	0.95	0.88	0.09	0.15
	sig var	3	0.95	0.90	0.09	0.14
Chocolate flavour	all variables	1	0.61	0.40	0.19	0.26
	sig var	3	0.79	0.38	0.14	0.26
Earthy flavour	all variables	1	0.74	0.53	0.14	0.21
	sig var	3	0.75	0.29	0.14	0.26
Spice flavour	all variables	1	0.63	0.51	0.16	0.21
	sig var	3	0.94	0.77	0.07	0.14

^† For the PLS model with all variables, Factor 1 explained 56% of the X variance and 57% of the Y variance; for the PLS model with a sub-set of significant variables, Factor 1 explained 84% of the X variance and 52% of the Y variance; Factor 2 explained a further 7% and 14% of the X and Y variance, respectively; ^‡ PLS model parameters where cal = calibration, val = validation, RSME = root mean square error of prediction; ^§ Sig var indicates a sub set of significant variables selected using an uncertainty test and high correlation loadings.