Model evaluations for individual DA estimates is presented.

KL designates the KL divergence between the kernel density estimate of measured and DA-output glucose, LC designates linear correlation, potentially with a lag. Notice: (i) all models produce reasonable results, the ultradian mean glucose forecasts are the most accurate, (ii) the accuracy of the continuous forecast is important because the forecasts are produced every minute, generating enough glucose values to render an accurate short term HbA1c forecast that patients can immediately relate to their nutrition intake, (iii) the metrics do not always select the same best model, (iv) no single model is the representative of all patients, and (v) both models are capable of representing and forecasting patients.