Re: Principle component analysis

[Oguz Yarimtepe]

On Thu, 2009-04-16 at 09:24 -0600, E. Joshua Rigler wrote:
> My guess is that the graphic you linked to is a plot of the first
> column of the second output argument to princomp.m (i.e., the
> "transformed data").  Either call princomp.m with a second output
> argument, or multiply the demeaned input matrix by the principal
> component matrix.  (e.g.,
> 
> >> pc = princomp(obs_matrix);
> >> xform_matrix = center(obs_matrix)*pc;
> 
> This is exactly what princomp.m does to get the transformed data.

I downloaded the princomp.m from svn and here is what i got:

test princomp.m
  ***** test
 x=[1,2,3;2,1,3]';
 [pc,z,w,Tsq]=princomp(x);
 m=[sqrt(2),sqrt(2);sqrt(2),-sqrt(2);-2*sqrt(2),0]/2;
 m(:,1) = m(:,1)*sign(pc(1,1));
 m(:,2) = m(:,2)*sign(pc(1,2));
!!!!! test failed
error: `zscore' undefined near line 20 column 29


I am using the octave3.0 from the Ubuntu repo. Should i checkout from
svn and install it?

-- 
Oguz Yarimtepe
http://www.loopbacking.info