Model Discrimination Design is a DOE strategy specifically designed to distinguish between competing statistical models — selecting experiments that maximize the expected difference between model predictions, enabling efficient determination of which model best describes the process.

How Model Discrimination Works

• Competing Models: Specify two or more candidate models (e.g., linear vs. quadratic, different interaction terms).
• T-Optimal: Find design points where the predicted responses from competing models differ maximally.
• Experiments: Run experiments at the discriminating points.
• Selection: Use model comparison criteria (AIC, BIC, F-test) to select the best model.

Why It Matters

• Efficient Resolution: Resolves model ambiguity with minimum additional experiments.
• Model Selection: Critical when data from an initial experiment doesn't clearly distinguish between models.
• Sequential: Often used as a follow-up to an initial response surface experiment.

Model Discrimination Design is letting the data choose the model — designing experiments specifically to reveal which mathematical model truly describes the process.