Fig. 3.

The iterative calibration process used for the parameter estimation in this study. At each iteration, the top five significant parameters are determined by the sensitivity analysis (SA) that consists of the fractional factorial design (FFD) and the analysis of variance (ANOVA). The significant parameters are sent to ABC for the parameter inference. At each iteration, the model is executed 5000 times with the parameter sets sampled from the pool of significant parameters. The posteriors are generated using the top 100 results. If the posterior is significantly narrower than the prior, the median of the posterior is accepted as the estimated value. If not, the parameter is added to the pool of the free parameters for the next round of calibration.