Translator Disclaimer
October 2020 A note on a Comparison of Starting Points for the Generation of D-optimal Second-order Designs
Ngozi Chizoma UMELO-IBEMERE
Afr. J. Appl. Stat. 7(2): 1001-1008 (October 2020). DOI: 10.16929/ajas/2020.1001.253

Abstract

Usually, line search algorithms begin by an initial guess of the starting point. However, for quick convergence and reduction in computational cost, it is necessary to determine an optimal starting point. In this paper, comparison of three starting points for the generation of second-order \(D-\) optimal designs using a variance-gradient based exchange algorithm is considered. Quantile plots are used to compare the variances of \(D-\)optimal designs generated by each of the three starting points. Result obtained reveals that an optimal choice of the starting point improves \(D-\)efficiency of the generated design points.

Les algorithmes de recherche en ligne commençent par un point donné et il est souhaitable de déterminer ce point de maniére optimale afin d'assurer une convergence plus rapide et de réduire le coût de calculs. Dans ce papier, nous comparons trois méthodes de détermination de trois points de départ pour la génération de second ordre \(D-\) dans un design optimal, utilisant une méthode de variance-gradient. Des études de simulations illustrent notre méthode.

Citation

Download Citation

Ngozi Chizoma UMELO-IBEMERE. "A note on a Comparison of Starting Points for the Generation of D-optimal Second-order Designs." Afr. J. Appl. Stat. 7 (2) 1001 - 1008, October 2020. https://doi.org/10.16929/ajas/2020.1001.253

Information

Published: October 2020
First available in Project Euclid: 15 January 2021

Digital Object Identifier: 10.16929/ajas/2020.1001.253

Subjects:
Primary: 40A05
Secondary: 62K05 , 62K20

Keywords: D-efficiency , line search method , quantile plots , second-order designs , starting points

Rights: Copyright © 2020 The Statistics and Probability African Society

JOURNAL ARTICLE
8 PAGES

This article is only available to subscribers.
It is not available for individual sale.
+ SAVE TO MY LIBRARY

SHARE
Vol.7 • No. 2 • October 2020
Back to Top