Uncertainty quantification

Uncertainty quantification studies how uncertainty in data, parameters, initial conditions and model structure affects conclusions and predictions.

Parameter uncertainty

Instead of using only one fitted parameter set, plausible parameter values can be propagated through the model to obtain a distribution or interval of possible outputs.

Stochastic uncertainty

In stochastic models, repeated simulations can differ even when parameters are fixed. This intrinsic variability should be distinguished from uncertainty about the parameter values themselves.

Structural uncertainty

Different reasonable model formulations may produce different conclusions. Comparing model structures can therefore be part of uncertainty assessment.