A Parametric Framework for the Comparison of Methods of Very Robust Regression

Marco Riani, Anthony C. Atkinson and Domenico Perrotta

- A .pdf file made up of three subsections. In the first we give a second motivating example. The second and third are expanded versions of our analyses of Examples 2 and 3 in the paper. The numbering of the figures continues that of the paper.
- An .avi file which shows the animation of Figure 2 of the paper.
- A .ppsx file which illustrates the concepts of: Design region, Theoretical overlapping index and Empirical overlapping index

Out MATLAB toolbox FSDA which has been used to generate all the results given in the paper can be downloaded from http://www.riani.it/MATLAB .

More specifically: the relevant functions inside FSDA which have been used are:

FSR.m = Forward search in linear regression

Sreg.m = S estimator in linear regression

MMreg = MM estimator in linear regression (Sreg is called in order to have a robust estimate of the scale)

MMregcore = MM regression estimators given an estimate of the scale.

LXS = LMS (Least median of squares) and LTS (Least trimmed squares)

The initial part of these files contains extensive documentation of the input/output arguments and a series of examples.

Additional documentation can be found in the html pages of FSDA integrated in the MATLAB documentation system (see screenshot below).