| If all models and wrong, but some are useful, then how does one discover which are the useful models? How does one discover, correct, or mitigate model error? How does one quantify the uncertainties in using an imperfect model? Is probability a satisfactory tool to describe uncertainty in this situation? I have no definitive answer to all these questions, but I will argued that dynamics is the key to understanding, and dealing with, model error. |
|