R help - Apr 2005 - Princomp$Scores

Hi all,

I was hoping that someone could verify this for me-

when I run princomp() on a matrix, it is my understanding that the scores 
slot of the output is a measure of how well each row correlates (for lack of 
a better word) with each principal component.

i.e. say I have a 300x6 log2 scaled matrix, and I run princomp(). I would 
get back a $scores slot that is also 300x6, where each value can be negative 
or positive. I'd assume that the negative values correspond to rows that are
negatively correlated with that particular PC, and vice-versa for positives.

Thanks in advance for the help,
Ken

The scores can be "manually" computed as 

    center(your.matrix,scale=F) %*% eigen(cov(your.matrix))$vector  # or cor()

i.e. a linear combination of your mean-corrected data. negative
coefficients does not imply anything, because computation of the signs
of eigenvectors is arbitrary. Absolute value of
princomp(your.matrix)$load indicates which variable influences a
principal component most (this is closer to the "correlate" thing you
are thinking of, I guess).

On Apr 9, 2005 1:09 AM, Ken Termiso <jerk_alert at hotmail.com>
wrote:
> Hi all,
> 
> I was hoping that someone could verify this for me-
> 
> when I run princomp() on a matrix, it is my understanding that the scores
> slot of the output is a measure of how well each row correlates (for lack
of
> a better word) with each principal component.
> 
> i.e. say I have a 300x6 log2 scaled matrix, and I run princomp(). I would
> get back a $scores slot that is also 300x6, where each value can be
negative
> or positive. I'd assume that the negative values correspond to rows
that are
> negatively correlated with that particular PC, and vice-versa for
positives.
> 
> Thanks in advance for the help,
> Ken
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>