|Date: ||Tue, 29 Apr 2008 14:15:25 -0500|
|Reply-To: ||Citam <citam.sasl@GMAIL.COM>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||Citam <citam.sasl@GMAIL.COM>|
|Subject: ||Re: Variable Transformation|
|Content-Type: ||text/plain; charset=ISO-8859-1|
On 4/29/08, Mathur, Rajat <Rajat.Mathur@diamondconsultants.com> wrote:
> I am creating some variable transformation for running a liner regression. I have a variable 'Money Spend', I wish to see if change in 'Money Spend' affects my dependent variable. So what transformation should I use?
> Also is there any guide book or rule of thumb for transformations to business problems?
Why might you want to transform the data?
Business problems are no different from other problems. First, you
should develop a model based on your knowledge that will test the
hypothesis of interest. Second you should examine the residuals to
see if they fit the assumptions of the model (independent, identically
distributed N(0,sigma)). Although transforming an independent
variable is possible, it is more customary to transform the dependent
variable if the assumptions are (grossly) violated. The online docs
provide a nice description of the procedures. You might want to
consider the ROBUSTREG, not to mention the experimental ODS output for
the statistical procedures, like REG.