15 may. 2013

Robust Mahalanobis Ellipse

In this plot we compare the Mahalanobis ellipse, using as center, the mean of the columns and as covariance, the classical covariance (is the same plot than in the previous post, but in this case I added colors and symbols to the samples, according to their sample set), and the Mahalanobis ellipse, based on the robust statistics (in this case the Minimum Covariance Determinant).
The calculation of the MCD is done with the function: covMcd( ). This function gives new robust values: center and cov to use indeed the classical ones.

We see how the distribution of the samples, in the ellipse, change, and also the number of outliers detected.

No hay comentarios:

Publicar un comentario