Prev: Hide wait bar
Next: colorbar lighting
From: Samoline1 Linke on 11 Nov 2009 04:47
Hi,

I have seen that a very good details is given over Factor Analysis and Principal Component Analysis in Matlab.

But I think they did not mention that these methods are not suitable for non-linear data. Could you please confirm that? Or can we apply the same factor analysis to nonlinear data as well?
From: Peter Perkins on 11 Nov 2009 10:29
Samoline1 Linke wrote:

> I have seen that a very good details is given over Factor Analysis and Principal Component Analysis in Matlab.
>
> But I think they did not mention that these methods are not suitable for non-linear data. Could you please confirm that? Or can we apply the same factor analysis to nonlinear data as well?

Factor Analysis, or at least the version implemented by FACTORAN, assumes the (linear) model given at the doc of the reference page

<http://www.mathworks.com/access/helpdesk/help/toolbox/stats/factoran.html>

I imagine there are papers in the literature discussing how robust the method is with respect to that assumption, but I am not familiar with them.
From: Samoline1 Linke on 12 Nov 2009 04:52
Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <hdel8o$3ne$1(a)fred.mathworks.com>...
> Samoline1 Linke wrote:
>
> > I have seen that a very good details is given over Factor Analysis and Principal Component Analysis in Matlab.
> >
> > But I think they did not mention that these methods are not suitable for non-linear data. Could you please confirm that? Or can we apply the same factor analysis to nonlinear data as well?
>
> Factor Analysis, or at least the version implemented by FACTORAN, assumes the (linear) model given at the doc of the reference page
>
> <http://www.mathworks.com/access/helpdesk/help/toolbox/stats/factoran.html>
>
> I imagine there are papers in the literature discussing how robust the method is with respect to that assumption, but I am not familiar with them.


ya I think it is not very robust against non-linearity. I could not get the logic but as it is based on comparing the correlations and the correlations strongly depend on whether ur data is linear or non linear so I think we have to take care of that......
 | 
Pages: 1
Prev: Hide wait bar
Next: colorbar lighting