FIG 1 .

Illustration of model checking without and with uncertainties. Following a range of experiments, standard data analysis highlights the distribution of values for a given parameter α. (A) Along a standard model-checking protocol, one assumes a single value for α, usually the mean. Such a value is then used for model calibration, which allows a simulation. Simulation results are then compared with observations for the sake of model verification (e.g., usually via linear regression between prediction and observations). (B) An example of a model-checking protocol that considers uncertainties per se. Instead of considering a single parameter value, one considers a range of values and precision guarantees and performs a range of simulations accordingly (one per color). Altogether, this SMC approach validates the models while taking into account intrinsic uncertainties and guarantees the desired precision (90% here).