Operation Research - Validation of a Model

...

Dynamic Validity: It is determined with maintaining how the model will be maintained during its life cycle so it will continue to be an acceptable representation of the system.2 areas which are of importance are updating and review. The model developers need to establish a procedure by which information is collected and analyzed to determine if and when model parameters need to be changed and a process by which such changes are affected. Also, a review procedure is required that is a regular schedule for reviewing the success or failure of model during its life cycle. The reviews must involve decision makers, solution implementers and model developers in sessions that cover divergences between predicted solution and actual outcomes, proposed model changes and new evaluation of model’s validity.

Q3. Describe the role of sensitivity analysis in testing the operational validity of a model?

A3. In conducting sensitivity analysis, values are systematically varied of the model parameters over some range of interest to see if and how the recommended solution changes. The model is run with extreme values of parameters and comparing the solution set of decision variables and their values with the recommended solution. If the solution is sensitive to certain parameter changes then the model should further analyse and justify in detail the values of the parameters. In case of an optimizing model, optimizing criterion may not be sensitive to large changes in the value of a decision variable and so the need for finely tuned parameters might be alleviated. It also involves the relationship between small changes in parameters values and magnitude of related changes in outputs.it helps to understand how to model with respect to real system. For complex models the implementation validity is important factor. The concern is the extent to which real world system being modeled responds in a manner indicated by the recommended model solution.

Q4. What does the author say about whether there is a validation technology that is appropriate for all models?

A4. There can be no model which can be appropriate for all models. There can be validations which are successful for models for which historical data can be used but for futuristic of first time models; validity is superseded by model credibility. Even in policy model myth exists that it can be fully validated as these models can at best be invalidated. First time models or nonexistent systems are validated differently than real systems.

Q5. Cite the page in the article that lists basic validation steps.

A5. a. Model Assessment- Page- 610

b. Model Validation- Page -611

c. Basic validation step- Page- 616

...