Construction of factorial-component orthogonal arrays

Component orthogonal arrays are experimental designs that have been recently introduced in statistical community with the aim of collecting intelligent data for analyzing order-of-addition experiments. In this paper, we propose a systematic method for constructing component orthogonal arrays with two-level components, and obtain a new type of designs called factorialcomponent orthogonal arrays. Such a design is optimal in terms of balance, as it ensures orthogonality between any pair of design columns. This characteristic allows for broad generality for empirical modeling. In addition, this design typically has much fewer experimental runs than the conventional design obtained by Cartesian product. The costeffectiveness of the proposed design is demonstrated by a real combination experiment of drug cocktails.