Optimizing the Design of an Experiment using the ADOpy Package: An Introduction and Tutorial

AbstractHow can researchers make experiments more efficient without sacrificing precision of measurement? Creative approaches, such as adaptive (staircase) methods in psychophysics, have been proposed as a solution. Advances in Bayesian statistics offer algorithm-based ways to achieve the same ends. The first part of this tutorial provides a conceptual introduction to one such approach, adaptive design optimization (ADO), along with examples of its application. The second part consists of a tutorial that introduces, in a hands-on training environment, an open-source Python package, ADOpy, that contains the ADO engine. The package was written using high-level modular-based commands such that users can use the package without having to understand the computational details of the ADO algorithm. The package is available from GitHub with three pre-installed experimental tasks in psychophysics, delay discounting, and risky choice. Converting a non-ADO task into an ADO-based task is straightforward, which will be demonstrated with working examples in PsychoPy.

Return to previous page