[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Re: Taking the log of variable with Positive,
From:       Doyle Kalumbi <djrk0003 () COMCAST ! NET>
Date:       2010-02-27 19:58:44
Message-ID: 1859083752.9075411267300724463.JavaMail.root () sz0053a ! emeryville ! ca ! mail ! comcast ! net
[Download RAW message or body]

Peter: 



Thank for your observation. There is an error in my definition of the variable: Yt is \
weight (lb) in current week    and Yt-1 is the weight   (lb)   in the prior week. 



Thanks, 



Doyle. 





----- Original Message ----- 
From: "Peter Flom" <peterflomconsulting@mindspring.com> 
To: djrk0003@COMCAST.NET, SAS-L@LISTSERV.UGA.EDU 
Sent: Saturday, February 27, 2010 5:44:48 AM GMT -08:00 US/Canada Pacific 
Subject: Re: Taking the log of variable with Positive, Zero and Negative Values 

Doyle Kalumbi <djrk0003@COMCAST.NET> wrote 
> Users: 
> 
> 
> 
> I have a variable that is defined as a weekly growth rate (WGrt)  defined as (( Yt- \
> Yt-1)/Yt-1)*100;   where Yt is Growth rate in the current week, Yt-1 is the Growth \
> rate in the prior week. Values for WGrt can be as high as +20 and can be as low as \
> -50 with zero   values  and   various positive and negative numbers in between. I \
> would like to estimate a linear model with both my dependent   and independent \
> variables being in log form so that my estimated coefficients can be interpretted \
> as elasticities. What is a statistical valid way of taking log(WGrt)   in such \
> circumstances   such that this new variable   has no distortions?  

Hmmmm. 

The WGrt is the rate of growth of a growth rate, right?   Like "acceleration"? 

the log of a ratio is the difference of the logs, but if some of the growth rates are \
negative, that's still a problem.   Still, you might want to use the difference of \
growth rates (on the regular scale) rather than the log of the ratio of rates. 

HTH 

Peter 

Peter L. Flom, PhD 
Statistical Consultant 
Website: http://www DOT statisticalanalysisconsulting DOT com/ 
Writing; http://www.associatedcontent.com/user/582880/peter_flom.html 
Twitter:    @peterflom 


[prev in list] [next in list] [prev in thread] [next in thread]