Date: Tue, 28 Aug 2007 15:16:22 -0700
Reply-To: "Noble, Lyndsay" <Lyndsay.Noble@ejgallo.com>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Noble, Lyndsay" <Lyndsay.Noble@ejgallo.com>
Subject: Principle Components Regression and Time Series Analysis
Content-Type: text/plain; charset="us-ascii"
I posted this a couple of months ago and received no response, so I'll try
again before spending multiple hours in syntax development.
I have heard through the grapevine that, several years ago, someone had
developed a macro (or syntax or script) that would run PCR in a time series
setting. That is, run a PCA on the data, use the principle components as
transfer functions in a time series analysis, and convert the coefficients of
the PCs into coefficients of the original independent variables.
Do any of you wonderful, helpful listers have any knowledge of this fabulous
tool? I can do the math myself, but the code would save me hours of time and
I would feel less likely to make a stupid algebra mistake.
Thanks in advance for any input and assistance.
Lyndsay C. Noble
Statistician, Information Intelligence