Comparative study between three methods of outlying detection on experimental results

This paper describes experimental results through multivariate statistical methods that might reveal outliers that are rarely taken into account by analysts. The results were submitted to three procedures to detect outliers: Mahalanobis distance, MD, cluster analysis, CA, and principal component analysis, PCA. The results showed that although CA is one of the procedures most often used to identify outliers, it can fail by not showing the samples that are easily identified as outliers by other methods, like MD. Mahalanobis distance proved to be the simpler application, with sensitive procedures to identify outliers in multivariate datasets.