23.6 Projection Models

The form of the model is very important in measuring variability

• e.g. simple trend vs time series; expected parameters vs pool of distⁿ for those parameters

Use common sense to ensure model is structurally consistent with the underlying process

Risk modeling we are more focused on the spread of potential outcomes instead of the best estimate

• Selecting models using parsimony will lead to unrealistically stable results

• Add flexibility (e.g. additional parameters) transfer assumptions from the structure to the parameters and allow the risk model reflect the uncertainty in the assumptions