Minimum Vector Variance Estimator in Outlier labeling of Multivariate Data: Application to HIV patient in Indonesia

An outlier is an observation whose pattern does not follow the majority of the data. Outliers in this study were characterized by extreme distance values, both very small and very large, exceeding the predetermined value. The method used in this research is Minimum Vector Variance (MVV) method because it has good computational efficiency and is robust against outliers. Based on the MVV algorithm applied to data on HIV patients in Indonesia in 2016-2018. The results showed that the MVV method produced more extreme distances than the Mahalanobis distance in labeling outliers. In the research data, it is found that there are 16 regions including outliers of the 34 observation used.