The model F-value of 15.85 implied the model is significant. There is only a 0.01% chance that an F-value this large could occur due to noise. Furthermore, the values of “Prob > F” less than 0.0500 exposed the significant model parameters. In this case parameters\(A\), \(B\), \(C\), BC, \(A^{2}\), \(B^{2}\) and \(C^{2}\)are significant model terms (Table 3.5). Still on parameter significance to the model build-up, inter-parameter model term BC and the intra-parameter model terms (also regarded as squared parameters) -\(A^{2}\), \(B^{2}\) and \(C^{2}\) were parameters that fully project the power of the model. In the ANOVA above, p- values greater than 0.1000 indicated model terms were not significant, hence other interactive factors such as AC and AB fall into this category.
Lack of fit is the variation of the data around the fitted model. If the model does not fit the actual response behaviour, this will be significant thus the model would not be used as a predictor of the response. On the other hand, pure error sum of squares is a sum of squares calculated for residuals of repeated runs only; used as a measure of experimental error. Table 3.5 displays the values of the sum of squares and the mean square of the pure error as 0.000 respectively. There is no p-value and f-value computed for model lack of fit and pure error tests. This is because there is no variation in the response values between replicated runs (design points).
Table 3.6 Correlation Functions