It is not possible to estimate with precision, many of the model parameters. Changes are subject to uncertainty and are often contingent on other changes. Users can reflect their uncertainty in the parameters passed to the model by, for example, providing an 80% confidence interval for a parameter. In other cases, users may be asked to indicate the frequency with which certain scenarios might occur. The model handles these uncertainty intervals or frequency distributions, with a Monte Carlo simulation. Each run of the simulation randomly samples a single value for each parameter from the uncertainty intervals or frequency distributions. The users can specify how many Monte Carlo simulations they wish to run. The model results take two forms. (1) a principal projection: the average (mean) of all the Monte Carlo simulations which have been run (2) a distribution of results, one for each of the Monte Carlo simulations, illustrating how key metrics might vary under alternative model parameters.
https://connect.strategyunitwm.nhs.uk/nhp/project_information/modelling_methodology/modelling_uncertainty.html#handling-uncertainty
Handling uncertainty
It is not possible to estimate with precision, many of the model parameters. Changes are subject to uncertainty and are often contingent on other changes. Users can reflect their uncertainty in the parameters passed to the model by, for example, providing an 80% confidence interval for a parameter. In other cases, users may be asked to indicate the frequency with which certain scenarios might occur. The model handles these uncertainty intervals or frequency distributions, with a Monte Carlo simulation. Each run of the simulation randomly samples a single value for each parameter from the uncertainty intervals or frequency distributions. The users can specify how many Monte Carlo simulations they wish to run. The model results take two forms. (1) a principal projection: the average (mean) of all the Monte Carlo simulations which have been run (2) a distribution of results, one for each of the Monte Carlo simulations, illustrating how key metrics might vary under alternative model parameters.