Model Validation

Download a pdf copy of this help file  here

See also: Model errors, Informal auditing, Model design introduction

This section discusses various methods that can be used to help validate the quality and predictive capabilities of a model. Some techniques can be carried out during a model's construction which will help ensure that the finished model is as free from errors and as accurate and useful as possible. Other techniques can only be executed at a future time when some of the model's predictions can be compared against what actually happened but one may nonetheless devise a plan to help facilitate that comparison.

Key points to consider are:

• Does the model meet management needs?

• Is the model free from errors?

• Are the model's predictions robust?

The following topics describe the methods we use to help answer these questions:

• Ensure model meets decision-makers requirements

• Comparing predictions against reality

• Informal auditing

• Checking units propagate correctly

• Check a model's behavior: View random scenarios on screen and check for credibility, Split up complex formulas (megaformulas), Compare with known answers, Analyzing outputs, Stressing parameter values

• Comparing results of alternative models

Read on:  Comparing predictions against reality