CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Optimal planning for ordered logit and ordered probit models Authors:  Xiaojian Xu - Brock University (Canada) [presenting]
Abstract: Both ordered logit and ordered probit models often serve as appropriate frameworks for statistical analysis when ordinal responses are involved. Although statistical inferences based on these ordinal regression models have been studied extensively in the literature, very few developments, have been done for optimizing the designs of experiments in the context of either ordered logit or ordered probit regression. We discuss optimal designs, particularly for ordered logit and ordered probit regression. Maximum likelihood estimation and D-optimality are adopted. Our resulting D-optimal designs are presented along with a comparison study that indicates D-optimal designs outperform their competitors. We also address the dependency issue of a design on the unknown parameters by an attainable two-stage design process. Furthermore, any assumed model form may not be accurate and lead to low efficiencies. Therefore, we also construct robust designs under the consideration of possible model departures.