Robust Regression Examples


Using the MVE and MCD Subroutines

The MVE subroutine computes the robust estimation of multivariate location and scatter, which are obtained by minimizing the volume of an ellipsoid that contains h points. The MCD subroutine is similar. It minimizes the determinant of the covariance matrix that is computed from h points. In general, the MCD subroutine is faster than the MVE subroutine.

You can use these robust locations and covariance matrices to detect multivariate outliers and leverage points. Both subroutines provide a table of robust distances.

The MVESCATTER and MCDSCATTER modules are used in these examples for plotting the results. These routines are in the RobustMC.sas file, which is contained in the SAS/IML sample library.