O jakości danych w kontekście obserwacji oddalonych w wielowymiarowej analizie regresji

The paper presents different definitions of outliers. We also collate selected outlier detection techniques, which represent very different approaches to outliers identification: classical univariate method embodied in boxplots, Andrews' curves, methods based on Cook's distance and Mahalonobis' distance, local outlier factor method, support vector machines. Moreover we empirically examine the agreement between the results of outlier detection methods on the benchmarking, real world dataset.