Seminar| Institute of Mathematical Sciences
Time: Monday, October 14th, 2024 , 16:00-17:00
Speaker: Yaping Wang, East China Normal University
Abstract: Mixture experiments analyze how changes in component proportions impact the response variable within the experimental region of a simplex. This paper introduces a new criterion, named the mean L1-distance (ML1D) criterion, for constructing uniform designs in mixture experiments. This criterion allows flexibility in point size and showcases a more uniform pattern within the experimental region. We also explore the optimal Scheffé-type simplex-lattice designs under the ML1D criterion. An interesting discovery is that the uniform mixture designs and the optimal Scheffé-type simplex-lattice designs are connected. For a two-component mixture design, these two types of designs are proven to be equivalent. For more than two-component mixture designs, numerical equivalences between the two designs are observed. These findings strengthen the rationale for users to adopt these designs in mixture experiments for modeling and prediction.