the level of the variable is increased, whereas a positive coefficient
indicates an increase in the response. Furthermore, a significant
interaction suggests that the level of one of the interactive
variables may increase while that of the other may decrease for a
constant value of the response. The lack of fit, which is used to
measure the fitness of models, exhibited no significant F value
(p>0.05) in terms of the response variable studied. The results, as
shown in Table 3, indicated that the models were accurate for
predicting those response variations. The following response
surface models, Eq. 2 to 7 were fitted to the response variable (Yi)
and three independent variables (X1, X2 and X3)