Generating Order-of-Addition Designs with Flexible Run Sizes

Abstract

In some industrial, chemical and biopharmaceutical studies, physical and/or chemical properties of the process outputs can be very different depending on the order in which materials are added. A series of trials for exploring the impact of sequentially adding the materials in various orders is called an order-of-addition experiment. In this talk, I will introduce a systematic method to generate efficient designs for these kinds of experiments. The main idea of the proposed method is to split the full design into several Latin squares, then a fractional design can be obtained by juxtaposing some selected Latin squares. A computer-assisted search procedure is first implemented to generate efficient designs with small to moderate run sizes. On the basis of computer-generated designs, a recursive method is then used to generate efficient designs with large run sizes. By applying the proposed method, a series of efficient designs with flexible run sizes is generated for practical applications.