Open Access
April 2018 Global convergence and ascent property of a cyclic algorithm used for statistical analysis of crash data
Issa Cherif Geraldo, Assi N'Gessan, Kossi Essona Gneyou
Afr. Stat. 13(2): 1631-1643 (April 2018). DOI: 10.16929/as/1631.125

Abstract

In this paper, we consider an estimation algorithm called cyclic iterative algorithm (CA) that is used in statistics to estimate the unknown vector parameter of a crash data model. We provide a theoretical proof of the global convergence of the CA that justifies the good numerical results obtained in early numerical studies of this algorithm. We also prove that the CA is an ascent algorithm, what ensures its numerical stability.

Dans cet article, nous considérons un algorithme d'estimation appelé algorithme cyclic iteratif (CA) utilisé en statistique pour estimer le vecteur paramètre inconnu d'une modèle pour les données d'accidents. Nous donnons une preuve th'eorique de la convergence globale du CA qui justifie les excellents résultats numériques obtenues dans les études numériques antérieures dudit algorithme. Nous prouvons aussi que le CA augmente la log-vraisemblance à chaque itération, ce qui assure sa stabilité numérique.

Citation

Download Citation

Issa Cherif Geraldo. Assi N'Gessan. Kossi Essona Gneyou. "Global convergence and ascent property of a cyclic algorithm used for statistical analysis of crash data." Afr. Stat. 13 (2) 1631 - 1643, April 2018. https://doi.org/10.16929/as/1631.125

Information

Published: April 2018
First available in Project Euclid: 7 June 2018

zbMATH: 06885664
MathSciNet: MR3811761
Digital Object Identifier: 10.16929/as/1631.125

Subjects:
Primary: 62-04 , 62F10 , 62H12 , 62P99

Keywords: cyclic algorithm , global convergence , iterative method , maximum likelihood , road safety

Rights: Copyright © 2018 The Statistics and Probability African Society

Vol.13 • No. 2 • April 2018
Back to Top