These differences in the response values across the combinations
tested are due to the effects of the product and process
variables. To characterize and quantify these effects, both
measured responses, rAA and %AAr data sets were separately
fitted into Equation (2) (or lower order equations) through RSM.
Response fit analyses analysis of variance (ANOVA) presented in
Table 4 revealed that only the individual linear effects of heating
temperature, exposure time, and initial AA concentration were
significant predictors of rAA in the heated SFJ. This measured
response significantly fitted into the linear model (Equation (5)),
with an F-value of 421.76. Moreover, ANOVA results showed that
the there is only 0.01% chance that the calculated F-value could
occur due to chance (P < 0.0001). The lack-of-fit ANOVA further
revealed that the data set for the measured rAA values did not
have significant lack-of-fit (P ¼ 0.1192) to Equation (5). Other indicators
of the goodness-of-fit of the rAA data set include the
coefficient of determination (R2) of 0.982, adjuster R2 (Adj. R2) of
0.980, and predicted R2 (Pred. R2) of 0.975. Frost (2013), Motulsky
(2014), and Kozak, Kozak, Staudhammer, and Watts (2008)
explained that the R2 value indicates how the data set fits in the
model, while the Adj. R2 value is a modification of R2 due to the
Fig. 1. Response surfaces showing the interactive influences of significant predictive intrinsic and process variables on rAA and %AAr.Center point values were used in the nondynamic
variables of