[R] cmdscale question

[Prof Brian Ripley]

Why not do your homework instead of sending the same message three times?

The references on the help page (especially Cox & Cox) will explain to you 
how scaling works on dissimilarities.

There are better alternatives for non-Euclidean dissimilarities: see MASS 
(the book) and its supporting software.

On Mon, 23 Jul 2007, mister_bluesman wrote:

>
> Hi.
>
> I know matrices that use distances between places works fine when using
> cmdscale. However, what about matricies such as:
>
>  A     B   C   D   E
> A 0    1   23  12  9
> B 1    0   10  12  3
> C 23  10   0   23  4
> D 12  12  23  0   21
> E  9   3    4   21   0
>
> i.e. matrices which do not represent physical distances between places (as
> they would not make sense for real distances such as the one above) but
> other statistics instead?
>
> Thanks
>
>

-- 
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