CFE 2017: Start Registration
View Submission - CMStatistics
B1355
Title: Outlier detection in multivariate data with robust Mahalanobis distance based on shrinkage estimators Authors:  Elisa Cabana - University Carlos III of Madrid (Spain) [presenting]
Henry Laniado Rodas - Universidad EAFIT Medellin (Colombia)
Rosa Lillo - Universidad Carlos III de Madrid (Spain)
Abstract: Different combinations of robust location and covariance matrix estimators based on the notion of Shrinkage are proposed. These collection defines robust Mahalanobis distances to address the problem of detecting outliers in multivariate data. The parameters needed for the shrinkage estimators defined, have been optimally estimated. The performance of the proposed distances is studied by means of a comparison with other existing methods from the literature, in simulated scenarios and with a real dataset example. The good computational results and the high correct detection rates and low false detection rates in the vast majority of cases, shows the advantages of our proposal.