If a catastrophe model produces highly unexpected loss estimates, what diagnostic steps do you take to validate the output?
