|
Activity Number:
|
354
- Experimental Design and Reliability
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #318742
|
|
|
Title:
|
Optimal Design Construction with Respect to Several Probability Distributions Using Clustering Approach
|
|
Author(s):
|
Muditha Lakmali * and Saumen Mandal
|
|
Companies:
|
University of Manitoba and University of Manitoba
|
|
Keywords:
|
Optimal Design;
Directional Derivatives;
Multiplicative Algorithm;
Probability Distributions
|
|
Abstract:
|
Optimization has been a basic mathematical tool in scientific research. The aim of our study is to apply optimal design theory to solve optimization problems with respect to several probability distributions and to improve the convergence of algorithms for obtaining optimal designs. When we run a multiplicative algorithm, the resulting design turns out to be a distribution of disjoint clusters of points which can be defined in terms of conditional distributions and a marginal distribution across the clusters. We transform this clustering approach to a general problem of optimization with respect to several distributions. The performance of our approach is investigated by constructing D-optimal and Ds-optimal designs for regression models including a practical model in Chemistry. Our approach shows considerable improvements in convergence of the algorithm. Using the properties of the directional derivatives of each criterion function, the convergence of the algorithm is improved. Results for each model will be reported, and applications in others areas will be discussed.
|
Authors who are presenting talks have a * after their name.
