ICAOD: An R Package for Finding Optimal designs for Nonlinear Statistical Models by Imperialist Competitive Algorithm

Area of Science:

• Statistics
• Computational Statistics
• Experimental Design

Background:

• Optimal design is crucial for cost-effective experimentation in various scientific disciplines.
• Existing tools for optimal design of nonlinear statistical models are limited, especially for non-differentiable criteria.
• There is a need for accessible and effective computational tools for generating diverse optimal designs.

Purpose of the Study:

• To introduce ICAOD, a novel R package for computing optimal experimental designs for general nonlinear statistical models.
• To provide a free and effective tool for generating locally, minimax, and Bayesian optimal designs.
• To demonstrate the utility of the imperialist competitive algorithm (ICA) for optimal design problems.

Main Methods:

• Development of the R package ICAOD.
• Implementation of the imperialist competitive algorithm (ICA), a metaheuristic optimization algorithm.
• Application of ICAOD to several case studies involving nonlinear statistical models and various optimal design criteria.
• Inclusion of theory-based tools for assessing design optimality.

Main Results:

• ICAOD successfully generates various types of optimal designs for nonlinear models.
• The imperialist competitive algorithm (ICA) proves effective for finding optimal designs, even with non-differentiable criteria.
• The package provides a practical and accessible solution for researchers needing optimal experimental designs.
• Demonstrated capability and effectiveness through multiple applications.

Conclusions:

• The ICAOD package offers a valuable, free resource for statisticians and researchers.
• The imperialist competitive algorithm (ICA) is a powerful tool for addressing complex optimal design challenges.
• ICAOD facilitates cost-effective and efficient experimental design in nonlinear modeling.