Extract bag plot filter does not use good criteria to choose outliers
This issue was created automatically from an original Mantis Issue. Further discussion may take place here.
The identification of outliers should not be based on a limit of the density of the data points but on a decreasing sorting in density of points.
In the Rainbow package of the R software, the algorithm used is the Hyndman one to identify high density regions.
Computing and Graphing Highest Density Regions, Rob J. Hyndman, The American Statistician, Vol. 50, No. 2 (May, 1996), pp. 120-126
The ACP_classique.R script is using a classical PCA and does not use the Rainbow package. So those the result can be directly compared to Paraview with default parameters.
The median curve is the number 15 et the only outlier is the 21. Those results can be used to build a simple unitary test. A more complicated one would compare the two PCA projections (execute the R script on the 2 first components).