RESULTS AND DISCUSSION 4.1 Fitting the model for the sensory parameters A summary of the statistics of the regressions fitted showed the standard deviation, the R2 and adjusted R2, predicted R2 and the predicted residual error sum of squares (PRESS). These were the indicators of how well the sensory models fitted the data. For goodness-of-fit, low standard deviations, R2 near 1 and relatively low PRESS were desired (Design-Expert, 2007). For colour (Table 5), the predicted R-squared of 0.36 was in reasonable agreement with the adjusted R-squared of 0.51 therefore the regression model for colour was significant (p= 0.0001<0.05). However, since there were negative predicted R-squared values for taste (-0.22), aroma (-0.01), spreadability (-0.13) and finger feel (-0.38), it implied their regression models were not significant (p>0.05) and this implied the overall means were better predictors of the sensory responses. According to Myers and Montgomery ( 2000), adequate precision measures the signal to noise ratio and a ratio greater than 4 is desirable; therefore the ratios of 9.31, 4.15, 5.45, 4.74 and 4.21 obtained for colour, taste, aroma, spreadability and finger feel respectively indicate an adequate signal suggesting that their models could be used to navigate the design space. Table 5: A summary of the statistics of the analysis of variance for the sensory parameters Standard deviation Mean PRESS RSquared Adj RSquared Pred RSquared Adeq Precision - 35 - Colour 0.11 0.50 0.98 0.62 0.51 0.36 9.31 Taste 0.15 0.56 1.96 0.28 0.08 -0.22 4.15 Aroma 0.11 0.53 1.09 0.41 0.24 -0.01 5.45 Spreadability 0.16 0.60 2.10 0.33 0.15 -0.13 4.74 Finger feel 0.11 0.70 1.11 0.19 -0.04 -0.38 4.21 Since the R2 for all the responses were significantly below 1 and the differences between the adjusted R2 and the predicted R2 were greater than 0.2 (Myers and Montgomery, 2000), it presupposes that the means of the sensory parameters were better predictors rather than their response regressions. Therefore it was not necessary to generate any response models but rather the mean plots for the sensory parameters. Table 6 describes the F-values and p-value prob >F of the sensory parameters. For colour, the model F-value of 5.78 implies the model is significant and that there is only a 0.01 % chance that a model F-value this large could occur due to noise. For the taste, however, the model F-value of 1.39 implies the model is not significant and there is a 19.2 % chance that a model F-value this large could occur due to noise. On the other hand, taking aroma into consideration, the model Fvalue of 2.43 implied the model is significant and that there is only a 1.1 % chance that a model F-value this large could occur due to noise. Similarly for spreadability, the model F-value of 1.77 implies there is a 7.14 % chance that a model F-value this large could occur due to noise. The model F-value of 0.81 for finger feel implies the model signal is not significant relative to the noise and that there is a 65.50 % chance that a model F-value this large could occur due to noise.
การแปล กรุณารอสักครู่..