First, I think you are talking about the princomp method of biplot, and
not a general biplot.
There are many different biplots for PCA, some incorrectly implemented.
The derivation is in MASS4 pp. 310-1, as well as the interpretation. You
need to understand what a biplot is to understand the units, which depend
on the value of 'scale' and 'pc.biplot'. See the note on the
help page
for biplot.princomp.
On Mon, 9 May 2005, Mathias Ditzen wrote:
> I am afraid this might be a stupid question, but after scanning a little
bit
> through the help archive and the internet, I didn't come up with an
answer
> myself.
> I have done a principle component analysis on a data set, which yields the
> mean response of 14 olfactory neurons to 16 different odors. When plotting
> the result in a biplot, it nicely shows me the odors in different positions
> in the principle component space. Additionally, the neurons are shown as
> vectors within this plot. My question now is: what are the units of the
axes?
> When doing the principle component in IDL, the relative positions of the
> odors are the same, but units on the axes span a much wider range.
> Additionally I couldn't figure out what the units in the variable space
> (right and upper border in the biplot) do represent.
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, 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