The Forward Search (FS) consists in ordering the observations by closeness to the multivariate normal model and progressively including them into the subset that is used for parameter estimation and that, at the beginning, is outlier free. In this paper we propose to enlarge the applicability of FS-oultier detection methods by assuming a mixture of K>1 normal components as a null model. Both the identification of the starting subset and the criterion for progressing in the search are based on the estimated values of the mixture density; outlying observations are detected by monitoring the values of a proper statistic during the search, as suggested by Atkinson et al. (2004)
Outlier detection via Forward Serach: a proposal based on mixture models / Calò D. G.. - STAMPA. - (2007), pp. 373-376. (Intervento presentato al convegno Meeting of the Classification and Data Analysis Group of the Italian Statistical Society tenutosi a Macerata nel 12-14 settembre 2007).
Outlier detection via Forward Serach: a proposal based on mixture models
CALO', DANIELA GIOVANNA
2007
Abstract
The Forward Search (FS) consists in ordering the observations by closeness to the multivariate normal model and progressively including them into the subset that is used for parameter estimation and that, at the beginning, is outlier free. In this paper we propose to enlarge the applicability of FS-oultier detection methods by assuming a mixture of K>1 normal components as a null model. Both the identification of the starting subset and the criterion for progressing in the search are based on the estimated values of the mixture density; outlying observations are detected by monitoring the values of a proper statistic during the search, as suggested by Atkinson et al. (2004)I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
