Une piste pour la recherche de la "valeur centrale optimale".Discussion autour de la robustesse et du comportement de la "médienne", combinaison de normes Lp

Abstract

In this paper, we present our research about the concept of central tendency or location. The meAdian is first presented as a robust location estimator, built using the mean and the median linear combination and their variance assessment by a bootstrap. Despite its theoretical statistical efficiency, the meAdian is tested on several empirical batches of data, often encountered during spatial analysis process such as spatial filtering. We then study the meAdian behaviour relating to the p value of the Lp-norms and conclude on a discussion and further researches developed to locate the “optimal center”. This paper completes the one previously published in Cybergeo Press, which describes some of the (ma)thematical aspects of the meAdian (Josselin & Ladiray, 2002a).

Keywords

• central tendency
• spatial filtering
• robustness
• bootstrap
• Lp-norm
• meadian