Model calibration assessment based on root-mean-squared error (RMSE), Akaike information criterion (AIC) and the frequency in which each model attained the minimum AIC (FR), and multi-dimensional measure.

Regarding the multi-dimensional measure, 4-D features, i.e., bias, standard error of residuals, trend of error over time, and trend of error over the BV range, as well as the Euclidean distance defined in Eq 24 are computed and compared between the two models. Results show that the refined model has significantly better performance in terms of RMSE and multi-dimensional measure. In addition, the models are comparable in terms of AIC, where among all 16 subjects, the frequency in which each model attained the minimum AIC is equal, i.e., 8. P-values are obtained using paired t-test. Underline bold numbers indicates significant difference between calibration performance of the two models.