From: Yukihiro Ishii <yukiasais_at_ybb.ne.jp>

Date: Thu 01 Sep 2005 - 00:18:16 EST

Date: Thu 01 Sep 2005 - 00:18:16 EST

Dear R users,

I have a data set of 25 cases with 150-160 explanatory variables(the number of which depends on what I choose from 200 odd digitalized spectrum strength numbers) and one dependent variable(a sensory test result). My natural choice is to work on a principal component analysis using the explanatory variables, thus enabling to characterize and describe the data space, and make a regression of the dependent variable on the principal components.

But a colleague of mine transposed the data matrix and, using the cases as the independent variables, explained the dependent variable in terms of the principal components he had. He changed obviously the score for the rotation. The analysis gave a plausible story. But I can't be sure of the physical meaning of it.

My colleague says that this method is common in the image analysis proper, which he specializes in.

Is there anyone who can comment on this matter. Venables & Ripley says something to the effect that either method will do, but the authors do not seem to give a specific example.

In my trade(chemistry), the data is commonly analyzed by the PLS(Partial Least Suare) method, which seems to give more or less the same result. Only the contribution of the PC's seems to be different.

I would appreciate any help. Thank you.

-- Yukihiro Ishii <yukiasais@ybb.ne.jp> ______________________________________________ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.htmlReceived on Thu Sep 01 00:29:21 2005

*
This archive was generated by hypermail 2.1.8
: Sun 23 Oct 2005 - 16:09:46 EST
*