partial least squares

[bassbuckeye]

Is anyone using or has written a VI that performs partial least squares linear regression?  If anyone has, I would greatly appreciate being able to take a look at it (picture of block diagram) or possibly use it.  If someone is willing to do this, I will privately send you my email so that I can obtain any pertinent information to be used in citing/using your work. Thank you for your help.    

[gokul_bits]

The block diagram plz.. my mail id : gokulavasan_2002@yahoo.com...

[GerdW]

Hello bassbucke,

what about this:
http://forums.ni.com/ni/board/message?board.id=170&message.id=181772&query.id=28863#M181772

[bassbuckeye]

Thank you for the replies.  The block diagram on the website does show a linear regression type fit, although I am not so sure what algorithms Labview uses to generate the fit stats in some of its Linear algebra VIs (ie. Partial Least Squares, SIMPLS, or regular least squares regression). I should have been more specific in saying that I need to develop an algorithm that conducts a partial least squares regression on observation (dependent) vector Y and Independent variable matrix X.  The important part here is that I need to have access to the various projection and scores generated during the course of a partial least squares regression.  As far as a block diagram is concerned, I don't have one yet.  I am fairly proficient with Labview, and was more interested in seeing how people were generating and managing the many score and projection vectors generated during PLS rather than the details of using Labviews linear algebra functions.   If someone has something like this, this is what I would like to take a look at. Thanks in advance.

[chilly_charly]

Statistical analysis is one of the LabVIEW weakest parts. You'll have difficulties to find a PCA vi. No wonder that PLS is still to be developped !..

Of course, I too would be extremelly happy to ear about  properly designed chemometric tools.

May be we could try to start someting ourself 😉

[bassbuckeye]

Yes, labview is fairly weak in this area, which is strange because the nature of data flow in labview really lends itself to Monte Carlo and Bootstrap techniques that can be incorporated into advanced chemometrics.  I would definitely be willing to toss some ideas around, although I am in a time crunch right now.

[bassbuckeye]

Well, I got fed up and just programmed a PLS VI.  It is quick and dirty programming, but I should have it cleaned up in a few days. This is a good jumping point for advanced chemometric techniques, so  if anyone would want to take a look at it or get the VI, let me know. Thanks for the input to those who responded.

[bassbuckeye]

Oh, I forgot to mention that this VI is for only 1 dependent variable.

[chilly_charly]

I would be happy to have a look. Can you share it on this forum ?

[bassbuckeye]

Sorry it took a while for me to respond.  I am still cleaning up the partial least squares VI.  It works, it is just that when I built it, I was in a hurry and was very, very sloppy in the layout.  I should have this done in a week or so, although this depends on how fast I can finish some other programs.  In the meantime, there are several high quality Mathcad tutorials posted on the web, several associated with Universities.  Googling "Partial Least Squares Mathcad" will give you plenty of hits.  Keep in mind that Labview and Mathcad treat matrices and vectors completely different.