R help - Jan 2004 - data structure problem

HI, there,
   I have a data set with special structure.
   It is in n*(5*p): n is the number of observations or data points
                     5*p is the matrix for each data point
   I'd like to conduct discriminant analysis to this data set. How could I
do? And where could I find related references to solve this problem?

Thanks a lot!


Jieping Zhao
PhD student in Bioinformatics, NCSU
Lab homepage: http://coltrane.gnets.ncsu.edu/index.html

Without more information on the context of the data, it's hard to say much
that will be useful.

One possibility is to treat the 5*p entries as 5*p variables, and apply the
commonly available discriminant tools to that.  Given more information, it
might be possible to do better.  As an example, one data set that has been
used as benchmark is the scanned images of hand-written digits.  Each digit
is encoded in a k x k matrix of values expressing the grayscale level of
each pixel (don't remember what k is).  A straight-forward way to train a
algorithm for pattern recognition is to treat the data as having kxk
variables.  However, smarter (but custom-built, rather than off-the-shelf)
algorithms can make use of the fact that the data is actually an image, and
possibly get better results.

Cheers,
Andy 
> From: Jieping
> 
> HI, there,
>    I have a data set with special structure.
>    It is in n*(5*p): n is the number of observations or data points
>                      5*p is the matrix for each data point
>    I'd like to conduct discriminant analysis to this data 
> set. How could I
> do? And where could I find related references to solve this problem?
> 
> Thanks a lot!
> 
> 
> Jieping Zhao
> PhD student in Bioinformatics, NCSU
> Lab homepage: http://coltrane.gnets.ncsu.edu/index.html
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}

If the data somewhat resembles multivariate Gaussian, I suppose one
possibility is to construct (by hand) something like LDA, but with the
covariance matrix constrained to be block-diagonal.  Just an idea.

Cheers,
Andy

> From: Jieping [mailto:jzhao at unity.ncsu.edu] 
> 
> my situtation is that each data point is made up of p 
> correlated 5-dimension
> vectors. Those 5 dimensions are orthogonal.
> Any suggestions will be appreciated!
> 
> JP
>
> From: Liaw, Andy [mailto:andy_liaw at merck.com]
> 
> Without more information on the context of the data, it's 
> hard to say much
> that will be useful.
> 
> One possibility is to treat the 5*p entries as 5*p variables, 
> and apply the
> commonly available discriminant tools to that.  Given more 
> information, it
> might be possible to do better.  As an example, one data set 
> that has been
> used as benchmark is the scanned images of hand-written 
> digits.  Each digit
> is encoded in a k x k matrix of values expressing the 
> grayscale level of
> each pixel (don't remember what k is).  A straight-forward 
> way to train a
> algorithm for pattern recognition is to treat the data as having kxk
> variables.  However, smarter (but custom-built, rather than 
> off-the-shelf)
> algorithms can make use of the fact that the data is actually 
> an image, and
> possibly get better results.
> 
> Cheers,
> Andy
> 
> > From: Jieping
> >
> > HI, there,
> >    I have a data set with special structure.
> >    It is in n*(5*p): n is the number of observations or data points
> >                      5*p is the matrix for each data point
> >    I'd like to conduct discriminant analysis to this data
> > set. How could I
> > do? And where could I find related references to solve this problem?
> >
> > Thanks a lot!
> >
> >
> > Jieping Zhao
> > PhD student in Bioinformatics, NCSU
> > Lab homepage: http://coltrane.gnets.ncsu.edu/index.html
> >
> 
> 
> --------------------------------------------------------------
> --------------
> --
> Notice:  This e-mail message, together with any attachments, contains
> information of Merck & Co., Inc. (One Merck Drive, Whitehouse 
> Station, New
> Jersey, USA 08889), and/or its affiliates (which may be known 
> outside the
> United States as Merck Frosst, Merck Sharp & Dohme or MSD) that may be
> confidential, proprietary copyrighted and/or legally 
> privileged, and is
> intended solely for the use of the individual or entity named on this
> message.
> If you are not the intended recipient, and have received this 
> message in
> error, please immediately return this by e-mail and then delete it.
> --------------------------------------------------------------
> --------------
> --
> 
> 
> 
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}