Stochastic optimization algorithms of a Bayesian design criterion for Bayesian parameter estimation of nonlinear...

Area of Science:

• Statistics
• Mathematical Modeling
• Experimental Design

Background:

• Bayesian estimation is crucial for parameter inference in nonlinear regression models.
• Optimizing experimental design maximizes the information gained from studies.
• Existing methods may not efficiently handle complex Bayesian design criteria.

Purpose of the Study:

• To propose and evaluate three novel stochastic algorithms for optimizing a Bayesian design criterion.
• To assess the efficiency of these algorithms in the context of nonlinear regression models.
• To apply the optimized design to estimate radioiodine thyroid uptake kinetics.

Main Methods:

• Development of three stochastic optimization algorithms: a stochastic simplex with adaptive sampling, Kiefer-Wolfowitz, and pseudogradient algorithms.
• Simulation study comparing algorithm efficiency for a nonlinear model with a discrete prior.
• Application to a real-world experimental design problem.

Main Results:

• The proposed stochastic algorithms effectively optimize the Bayesian design criterion.
• Comparative simulations demonstrate the relative efficiencies of the algorithms.
• Successful application in designing an experiment for thyroid uptake kinetics estimation.

Conclusions:

• Stochastic algorithms offer a viable approach for optimizing Bayesian experimental design.
• The presented methods provide efficient tools for parameter estimation in nonlinear models.
• This work facilitates improved experimental design in fields like endocrinology.