23.5 Model Risk

When considering different distributions, need to adjust the log likelihood with penalty for the number of parameters

• Hannan-Quinn Information Criterion (HQIC) is recommended because it is a compromise of other information criteria which add large or smaller penalties for # of parameter

Create a pool of distⁿ that are consider reasonable:

1. Randomly select a mean for the parameters and a \(\mathbf{\Sigma}\) from the pool of distⁿ

2. Randomly draw the parameters based on the selected distⁿ

3. For each claim that is simulated, draw from the distribution with parameters from (2)