Two levels (low and high) were selected for each of the independent variables; the 24 factorial design that was used for briquette production is shown in columns 2 to 5 of Table 4, which also show the measured responses.
Since the effect of water in the binder was confounded with that of the starch in this experiment, additional experimental runs were conducted to assess the effect of water on the response variables, with and without starch. The responses for briquettes containing water only (without starch) are presented in rows 21 to 24 of Table 4, for comparison with the results for otherwise similar briquettes produced with both water and starch mixture in rows 9 to 12. In the statistical analysis, the effect of dry starch on briquette responses was assumed to be the same as when only rice husk and corn cob residues were used (i.e., with no water or starch in rows 17 to 20 of Table 4).
Statistical effects of variables and their interactions on the responses were calculated based on the individual replicate results shown in columns 7, 9 and 11 of Table 4[42]. Effects of the variables and interactions between the variables on a response are estimated as the differences between the averages for the high and low levels of a variable or interaction, and the total mean response. The highest order interactions of variables were assumed to be largely due to random noise [42]. Normal probability plots of the effects can be used to visualize the significance of the effects of individual variables on the responses [42]. The estimated effects can be read from the abscissa, against the standard deviation of the normal distribution on the ordinate. The scale of the ordinate has been adjusted such that a normal distribution appears as a straight line, i.e., points that lie on the straight line may be a result of normal random variability, whereas those that deviate from the straight line indicate significant effects of these variables or interactions on the response. Analysis of variance was also used to determine the statistical significance of the observed effects [42].
The fitted model for the predicted responses is shown as Eq. (2)[42], and Eq. (3) was used to calculate the residuals (ɛ) of the responses.
equation(2)
View the MathML source
Turn MathJax on
equation(3)
View the MathML source
Turn MathJax on
Where;
View the MathML source is the grand mean for each set of response data (e.g., green density)
j1, j2 … jn is the observed main or interaction effect of the variables
x1, x2 … xn is the respective sign of the observed effects for each response value.
A normal probability plot of the residuals was used to visualize the normality and check that all effects other than those included in the model are explained by random noise.