Fig. 9.

Estimation of cortisol model parameters for a noiseless data sampled at 1 min intervals, b data with with 20% proportional error sampled at 1 min intervals, c noiseless data sparsely sampled at 45 min intervals. The model was simulated for 8640 min (6 days) duration to observe the rate of parameter convergence. The noiseless case a resulted in parameter convergence closer to nominal values. The noisy b and sparsely sampled c cases resulted in fluctuation of the parameters about their nominal values, for which parameters were extracted by calculating the median of these fluctuating values. The noisy data were filtered by wavelet denoising. We densified the sparsely sampled data using the pchip function in MATLAB with output sampled at 5 min intervals