Speaker: 杨建峰，南开大学

Inviter: 熊世峰

Title: Robust Designs for Order-of-Addition Experiments Based on Swap Distance

Language: Chinese

Time & Venue: 2025.11.29 11:30-12:00 腾讯会议：979-347-585

Abstract: Order-of-addition experiments play a crucial role in various fields, enabling researchers to explore complex effects of components. A substantial amount of research on order-of-addition experiments assumes that the underlying model is known. However, in most practical cases, we do not know the true form of the model. The optimal design for a particular model may underperform when the model is misspecified. Therefore, a design with excellent space-filling properties emerges as a robust and reliable choice. This paper introduces the swap distance as a specialized metric to measure the dissimilarity between two points in an order-of-addition experimental design. An upper bound of the minimum swap distance is established for designs with any number of rows and components. This upper bound serves as a benchmark for assessing design efficiency. Furthermore, we propose several construction methods for maximin swap distance designs. Some of the constructed designs achieve the established upper bound while others are theoretically guaranteed to have a notably large minimum swap distance. To verify the robustness of the proposed designs, we conduct a case study for the job scheduling problem. The results demonstrate that the proposed maximin swap distance design significantly outperforms alternative designs, including those based on other distance metrics.