[R] MASS library rob.cov ellipse

[Prof Brian Ripley]

On Mon, 2 Jul 2007, Jenifer Larson-Hall wrote:

> Hi,
> I used to get a really useful graph when I ran the following command
> using the MASS library:
>
>> cov.rob(cbind(dekeyser$AGE,dekeyser$GJTSCORE),cor=T)
>
> Besides the regular output, a graph appeared that had the classical
> correlation and the robust correlation, and two ellipses, one
> surrounding the data that would be used in the classical correlation and
> the other surrounding the data in the robust correlation.

But all the data are 'used in the classical correlation'.  Maybe you mean 
a tolerance ellipse for the implied bivariate normal distribution?

> I've searched through the MASS library but don't see a separate command
> that could produce this graph. Does anyone know whether one exists, or
> did the graph just disappear in the newer version of R?

It was never in the MASS *package*.  cov.rob() is mainly intended for 
multidimensional problems, not 2D ones.   It is possible you were using 
methods from package 'robustbase', such as covPlot?


The author of cov.rob.


> Thanks for any help,

> Dr. Jenifer Larson-Hall
> Assistant Professor of Linguistics
> University of North Texas
> (940)369-8950

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595