Date: Mon, 23 Apr 2007 17:44:59 -0400
Reply-To: Kevin Roland Viel <kviel@EMORY.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Kevin Roland Viel <kviel@EMORY.EDU>
Subject: Re: Explaining a Log Transformation on a dependent variable?
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Mon, 23 Apr 2007, Greg wrote:
> This isn't a specific SAS question but I thought I askit here. I am
> new to
> multiple regression and unfortunitly had to do a log
> transformation on the dependent variable to fix normality issues. My
> problem now is trying to explain the meaining of the final model/
> formula with the log function. My final model has three independent
> varaibles as predictors (they were centered) and conceptually I am
> stuck and not able to explain my model with the log transformation.
> Could anyone kind of walk through a explanation? Model for discusion
> purposes: Y=10 + .005W + .07H + 1.2P.
How about something from the "source"?
"For ease of interpretation, we present the results of the regression
analyses as the estimated percent difference in the geometric mean FVIII
level among the exposed compared to the unexposed. This percent difference
is estimated by 100%*(exp()-1), when a log-transformation has been
used. For a continuous exposure, such as the depression score, the
expression 100%*(exp()-1) is interpretable as the difference, expressed as
a percent, in geometric mean FVIII:C level among those with a given
depression score compared with the corresponding mean among those with a
score that is one unit lower."
1. Flanders, W.D., R. DerSimonian, and D.S. Freedman, Interpretation
of linear regression models that include transformations or interaction
terms. Ann Epidemiol, 1992. 2(5): p. 735-44.
I say from the source because this was more or less crafted by Dr.
Flanders for a paper in my dissertation (not yet submitted).
Department of Epidemiology
Rollins School of Public Health
Atlanta, GA 30322