Orthogonal arrays
Orthogonal arrays are the combinatorial objects behind many statistical designs of experiments: tables in which every pair, triple or quadruple of columns contains all combinations of levels equally often. With Eric Schoen, Alan Vazquez, Peter Goos and others I work on the complete enumeration of such designs, and on the Orthogonal Array package (OApackage) that generates and analyses orthogonal arrays, optimal designs and conference designs.
A mixed-level orthogonal array with 16 runs
Some results:
- Complete enumeration of pure-level and mixed-level orthogonal arrays (J. Combin. Des. 2010); the resulting complete series of orthogonal arrays are available online
- Two-level designs to estimate all main effects and two-factor interactions (Technometrics 2017)
- Classification and systematic enumeration of definitive screening designs (Ann. Statist. 2019, Stat. Comput. 2022)
- Systematic enumeration of two-level even-odd designs of strength 3 (CSDA 2023)
- A divisibility theorem for the odd J-characteristics of two-level designs (arXiv 2026), conjectured during the even-odd enumeration and formally verified in Lean; see the blog post
- OApackage: C++ library with a Python interface, on PyPI with documentation (JOSS 2019)