Date: Tue, 21 Nov 2000 15:15:54 -0500
Reply-To: Tony Baglioni <ABaglioni@QUALCHOICEVA.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Tony Baglioni <ABaglioni@QUALCHOICEVA.COM>
Subject: Use of logs in regression
I'm not sure I can present this question clearly but here goes. I will be
doing some simple multiple regressions with percent of change in price as
the outcome measure. Because I am using percents and I am reasonably sure
that there is nothing near normal about the data, I will do a log
transformation of the outcome measure. What is the best way to evaluate the
residuals in terms of the original metric? I was thinking of just taking
the anti-log but I'm not sure that's appropriate. Anyone have any ideas?
A. J. Baglioni, jr., Ph.D.
Medical Data Analyst
QualChoice of Virginia