Using Historical Data to Model Uncertainty
Here, we looked at how we could come up with the right functions to model the uncertainty of key input variables.
Uncertainty is not fixed to an average assumed value. Rather, it varies. We are now trying to see how the uncertainty varies.
They key is to estimate the probability of distribution function for critical inputs to a simulation model. The way to do that is by analysing historical data.
Generally speaking, simulation models do not use historical data directly. The idea is to use the data to create a so-called probability distribution function for each variable and then use the function for the simulation model. The process of identifying an appropriate probability distribution function for a variable for which we have historical data is called Distribution Fitting. The analytics of a platform includes a distribution fitting tool.
Uncertainty is not fixed to an average assumed value. Rather, it varies. We are now trying to see how the uncertainty varies.
They key is to estimate the probability of distribution function for critical inputs to a simulation model. The way to do that is by analysing historical data.
Generally speaking, simulation models do not use historical data directly. The idea is to use the data to create a so-called probability distribution function for each variable and then use the function for the simulation model. The process of identifying an appropriate probability distribution function for a variable for which we have historical data is called Distribution Fitting. The analytics of a platform includes a distribution fitting tool.