A1243
Title: Construction and analysis of (near-)CCOP design for mixed-level order-of-addition experiments
Authors: Ching-Yen Chuang - National Tsing Hua University (Taiwan)
Frederick Kin Hing Phoa - Academia Sinica (Taiwan) [presenting]
Abstract: Mixed-level order-of-addition (OofA) experiments arise in settings where responses depend jointly on the sequence in which experimental components are added and on the levels at which these components are set, with different components potentially having different numbers of levels. In many applications, some components contribute only through their order, whereas others possess additional levels that further influence the response. Identifying optimal order level combinations is therefore a fundamental yet challenging task, as the number of possible configurations grows combinatorically and experimental trials are often expensive. To address this challenge, an extension of complete consecutive order-pairing (CCOP) designs to mixed-level OofA experiments is proposed. The proposed designs ensure complete coverage of all consecutive component pairs across levels while achieving minimal or near-minimal run sizes. Explicit construction methods are provided, structural properties of the designs are established, and their optimality with respect to run size under broad conditions is proved. For analysis, a procedure based on the distance-based linear model (DBLM) is developed, which accommodates both ordering effects and level effects within a unified framework. Numerical studies and a real drug-combination experiment demonstrate that the proposed design analysis strategy reliably identifies superior order level settings using substantially fewer runs than competing methods.
