C0 = Intercept coefficient; C1 = slope coefficient; Err[Ci] = standard error of coefficient ‘Ci’; p -value = probability Ci ≠ 0 (< 0.05); MAD = mean absolute deviation; RMSD = root-mean-squared deviation. See Table 1 and 2 for abbreviations.
a Deviations between IPR and IPA; calculated from Eq. 10.
b Deviations between IPR and IPA; calculated from Eq. 11.
coefficients (C1) were significant with values for four of the FAME being in the range 0.95-0.99, which was close to the desired value of unity (1) for a model confirmation analysis. The results
for MeC18:2 yielded a lower C1 coefficient (0.87), suggesting that its model equation under predicted IPA, with respect to its corresponding experimental IPRvalues. This was confirmed for four of the six data pairs evaluated for MeC18:2, specifically the IPR data at T = 50, 60, 90 and 100 °C (323.15, 333.15, 363.15 and 373.15 K).
The MAD and RMSD results obtained from application of Eqs. 10 and 11 to (IPR, IPA) data pairs are summarized inTable 3 . These data showed generally low values where MAD = 0.12-0.63 and RMSD = 0.19-0.98. MeC18:2 had the highest RMSD value, a result that may have been influenced by the relatively low C1 coefficient in its model confirmation equation. On the other hand, CaME yielded the highest MAD value (0.63), despite having a relatively high C1 coefficient (0.99) while the MAD value of MeC18:2 was slightly lower.
The residuals (IPR − IPA) calculated from applying the Model A type equations for the five FAME are presented in Figure 3 . In general, the residuals were ≤ 0.51 for 20 of the 26 total data pairs. Of the remaining six data pairs, only one residual exceeded 1.43, which was for