“…Also given are two measures of the fit of the data: the correlation coefficient of the Arrhenius plot (r 2 ) and the root-mean-square error of the individual model predictions with respect to the actual measured value of S at any given time, effectively the coefficient of variation (COV) of S. Perhaps the most striking aspect is the generally extremely high values of r 2 associated with the Arrhenius plots; always P0.95 and in most cases >0.99. Contrast this with the assertion by Bank [31] that any value over 0.8 is 'valid' (implying that many values Table 2 Model fit results for all data sets Designation ÀE k /kJ mol À1 (r 2 ), k 0 /day À1 (COV) ÀE d /kJ mol À1 (r 2 ), k 0 /day À1 , (COV) ÀE n /kJ mol À1 (r 2 ), k 0 /day À1 (COV) n Cross-ply E-glass/epoxy laminates VT1 120 derived using other modelling procedures are a good deal less robust). The error in predicting the mean value of S is also relatively low, $1-2% in most cases, indicating that the model is the 'right shape' (note that the typical COV for a single strength data point is >5% for most GRP data sets and >10% for most GRC data sets).…”