[R] PCA prcomp problem

[Prof Brian Ripley]

On Tue, 9 Nov 2004, Deirdre Toher wrote:

> I've just starting using the prcomp function, and I want to be able to extract
> individual principal components (e.g. PC1, PC2) in vector format.  I haven't
> been able to find any documentation that explains how to do this (or even if it
> is possible).  Any help on the subject would be greatly appreciated.

Depends of course what you mean: do you want the coefficients or the data 
predicted onto principal components?

The coefficients are component `rotation' of the result (see the help 
page).

Projections onto the PCs is component `x' of the result (see the help 
page).  E.g.

fit <- prcomp(USArrests, scale = TRUE, ret.x = TRUE)
PC1 <- fit$x[, 1]
PC2 <- fit$x[, 2]

OR predict(fit)  gives the projections of old (or new, with newdata) 
data onto the PCs.


-- 
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