CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1611
Title: Bayesian and maximin A-optimal designs for spline regression models Authors:  Julie Zhou - University of Victoria (Canada) [presenting]
Isaac Rankin - University of Victoria (Canada)
Abstract: Optimal designs for spline regression models with multiple unknown knots are studied using an A-optimality criterion. Locally A-optimal designs are constructed, which depend on the true values of the knots. However, in practice, the knots are never known exactly, but it may be reasonable to assume a prior distribution for these knots. Using the prior distribution, we apply a Bayesian or maximin efficiency criterion to construct optimal designs. Several theoretical results are derived. We also propose to use a numerical method for computing the optimal designs. The key to using the numerical method is to transform the design problems into convex optimization problems, and the method is very fast to compute optimal designs on discrete design spaces. Examples of Bayesian A-optimal and maximin efficiency optimal designs are presented.