Hi Alex,

If you are talking about pearson correlation, it may be not quite useful in your case where y in {0, 1}.

HTH

Alex Murphy wrote: > Hi, > > What I meant was that I noticed that the correlation between the transforme= > d > variables with y variable is noticeably stronger. > > Regards, > Murphy > > > On Tue, Mar 10, 2009 at 12:06 AM, Shiling Zhang <shiling99@yahoo.com> wrote= > : > > > > I noticed that many people in the credit scorecard industry uses WOE > > > transformations. Are there any statistical basis to this transformation= > ? > > > > The transformations is motivated by > > 1) Kullback-Leibler=92s divergence > > 2) regression requires that logodds is linear in xbeta or x*beta > > > > > > >I noticed that even for mere variables with no correlation can be > > transformed > > > into one that has excellent correlation. > > > > I am not sure what do you mean in this statement. > > > > But I don't think the correlation among independent variables will be > > reduced or inforced after trandformations. > > > > HTH > > > > > > > > > > Alex Murphy wrote: > > > Hi to all, > > > > > > I noticed that many people in the credit scorecard industry uses WOE > > > transformations. Are there any statistical basis to this transformation= > ? > > I > > > noticed that even for mere variables with no correlation can be > > transformed > > > into one that has excellent correlation. > > > > > > -- > > > Regards, > > > Murphy Choy > > > > > > Certified Advanced Programmer for SAS V9 > > > Certified Basic Programmer for SAS V9 > > > DataShaping Certified SAS Professional > > > > > > --=20 > Regards, > Murphy Choy > > Certified Advanced Programmer for SAS V9 > Certified Basic Programmer for SAS V9 > DataShaping Certified SAS Professional