A design criterion for symmetric model discrimination based on flexible nominal sets

Area of Science:

• Statistics
• Experimental Design
• Biophysics

Background:

• Model discrimination in experimental design typically assumes prior knowledge of the true model, which contradicts the experimental objective.
• Existing methods to overcome this limitation include symmetric techniques, Bayesian, minimax, and sequential approaches.

Purpose of the Study:

• To develop a genuinely symmetric criterion for experimental design that does not require prior knowledge of the true model.
• To introduce a computationally efficient method for model discrimination.

Main Methods:

• A novel symmetric criterion is proposed, utilizing a linearized distance between mean-value surfaces.
• Flexible nominal sets are introduced as a new tool for this approach.
• Monte Carlo evaluation using the likelihood ratio assesses discrimination performance.

Main Results:

• The proposed criterion demonstrates computational efficiency.
• The method effectively evaluates the discrimination performance between competing models.
• Successful application to a pair of enzyme kinetics models is presented.

Conclusions:

• The developed symmetric criterion offers a robust and efficient approach to model discrimination in experimental design.
• This method removes the need for prior assumptions about the true model, advancing experimental design capabilities.
• The approach has practical implications, particularly in fields like enzyme kinetics.