Model Error and Estimation Error

Last Friday, William L. Cooper came to give us a speech. In his speech, he analysize the impact interaction between model error and estimation error.

Usually, OR people aren't willing to think a question: WHAT IF MY MODEL IS WRONG. Though, we will consider WHAT IF THE ESTIMATION OF PARAMETERS ARE WRONG. So we will do all kinds of sensitivity analysis.

But in Cooper's model, it isn't enough. First he assumes we already have 'true model' vs 'wrong model'. Then he analysizes the situations where
two companies are both smart enough to use wrong model
two companies both use wrong model
one companies uses smart model and the other one uses wrong one
During the competition, two companies both need to estimate the value of parameters. The interaction between model error and estimation error makes the problem hard especially when you want to develop some stylized results.

An interestion question during the Q&A section, 'How do you know what's the true model'. Maybe you can list out all possible true models and analysize your employed model's performance under those true models.

We always solve a problem by creating a new problem.