Classification, clustering, and phylogeny estimation
Thu, 26 May 2005 11:55:31 -0400
Interestingly, L. L. Thurstone used this issue (Thurstone, 1947) to
illustrate the uses of Factor Analysis -- he introduced a data set that
consisted of the intercorrelations of many different possible measures
of the "size" of a set of cardboard boxes: e.g. volume, sum of edges,
Then he showed that you can extract factors from the data that can be
related to box height, width, and depth.
James E. Corter
Associate Professor of Statistics and Education
Chair, Department of Human Development
Teachers College, Columbia University
525 W. 120th St.
New York, NY 10027
Email: [log in to unmask]
Tel: (212) 678-3843
FAX: (212) 678-3837
>That choice is "unusual" because most workers do not use that method.
One reason >is probably that one would not measure the size of a box by
adding its length and >width - most would rather multiply them together.
>The problem with "size" is that simple word hides a large diversity of
possible >measures. One cannot decide what is size without specifying
some model and what >desired properties the measure of size is supposed
to optimize. One should not use >some measure just because it seems
"reasonable" intuitively. As I said previously, it >is fortunate that in
most biological applications the various choices are usually highly
>F. James Rohlf, Distinguished Professor & Graduate Program Director
State >University of New York, Stony Brook, NY 11794-5245