Back to PLS Helpwhy mean-centering is effectively like defining contrasts?
Posted on 05/05/10 05:28:29
Number of posts: 13
Dear List,
I couldn't really figure out why centering with respect to grand mean is effectively like defining contrast. I will give you an example: let's say that we have a raw matrix like A= [5,3,1] for which the grand mean would be GM=3. Now, if we conduct a mean centering on matrix A , we will get B=[2,0,-2].Up to here, everything is clear. but then lets see the result of cross-covariance between A and B which produce C=[-4,0,8,0,-4] which is not an orthogonal rotation of the mean-centered matrix.Do I understand something wrong?Can anybody provide me a numerical example of this?
Thanks,
Kami
I couldn't really figure out why centering with respect to grand mean is effectively like defining contrast. I will give you an example: let's say that we have a raw matrix like A= [5,3,1] for which the grand mean would be GM=3. Now, if we conduct a mean centering on matrix A , we will get B=[2,0,-2].Up to here, everything is clear. but then lets see the result of cross-covariance between A and B which produce C=[-4,0,8,0,-4] which is not an orthogonal rotation of the mean-centered matrix.Do I understand something wrong?Can anybody provide me a numerical example of this?
Thanks,
Kami
Replies:
Untitled Post
I'm Online
Posted on 05/05/10 08:45:04
Number of posts: 229
Dear Kami...
the math is spelled out in the 2004 Neuroimage paper (NeuroImage 23 (2004) S250–S263) - available in the "Journal Articles" section of the PLS site.
cheers
nancy
the math is spelled out in the 2004 Neuroimage paper (NeuroImage 23 (2004) S250–S263) - available in the "Journal Articles" section of the PLS site.
cheers
nancy
Login to reply to this topic.
