ADOpy: a python package for adaptive design optimization.

Area of Science:

• Experimental Psychology
• Computational Statistics
• Machine Learning

Background:

• Formal methods for identifying optimal experimental designs are lacking.
• Advances in Bayesian statistics and machine learning offer algorithmic approaches to design optimization.
• Adaptive Design Optimization (ADO) is a method to maximize data collection efficiency and informativeness.

Purpose of the Study:

• To introduce ADOpy, an open-source Python package for Adaptive Design Optimization.
• To lower the barrier to entry for using ADO in experimental research.
• To demonstrate ADOpy's utility through practical examples and simulations.

Main Methods:

• Implementation of ADO algorithm in a modular Python package (ADOpy).
• Tutorial introduction to ADOpy and the underlying ADO methodology.
• Walk-through examples in psychometric function estimation, delay discounting, and risky choice tasks.

Main Results:

• ADOpy provides accessible implementation of ADO for experimental design optimization.
• Simulations show ADO designs can be more efficient than random or staircase designs.
• The package facilitates rapid information accumulation with fewer trials.

Conclusions:

• ADOpy democratizes the use of advanced adaptive experimental design methods.
• The package enables researchers to improve inference and data collection efficiency.
• ADOpy supports broader adoption of adaptive design optimization in scientific research.