Chikubu et al. (1985) proposed the palatability estimation formula of rice by using sensory overall acceptance and 21 physicochemical properties of 34 varieties. The palatability estimation formula consists of five variables: protein content, maximum viscosity, minimum viscosity, breakdown, and starch-iodine blue value. Breakdown measured with a Brabender Visco-Amylograph is defined as the difference between maximum and minimum viscosities. Therefore the formula has a multicollinearity problem. The best regression model using Chikubu et al. data was determined by the AIC statistic, multicollinearity detection, and cross-validation. The best model was as follows: y = -0.165 x
3 -0.00424 x
7 +0.120 x
12 -8.84 x
14 -O.313 (R
2 =0.695), where y is the sensory overall acceptance, x
3 is the protein content, x, is the minimum viscosity, x
12 is the expanded volume of the cooking quality test, and x
14 is the starch-iodine blue value. The correlation of the calibration data set using 34 varieties between the best model predicted values and the actual values was 0.83, and that of the cross-validation data set removing one from 34 varieties was 0.75.
View full abstract