Figure 6. Many unique parameter sets describe the training data equally well.

(a) Comparison across 8 optimized parameter sets identified in Module 1 (case study) that each yield R² > 0.99. Simulated data generated with each of these parameter sets are shown in shades of gray (simulations overlap). (b) Distribution of each optimized parameter value. Each data point is an optimized parameter value. Most parameter values vary widely across multiple orders of magnitude while retaining very similar R² values. (c) Examples of cost function landscapes exhibiting parameter unidentifiability. A parameter is considered unidentifiable if it cannot be uniquely estimated.