Date:         Mon, 4 Aug 2008 10:13:38 -0700
Reply-To:     "sophe88@yahoo.com" <sophe88@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "sophe88@yahoo.com" <sophe88@YAHOO.COM>
Organization: http://groups.google.com
Subject:      which transformation
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hi,

Binary 0/1 logistic regression. A continuous variable (unique values
>30) has pretty good univariate correlation with the dependant
variable, ~5%missing only. It has been selected from about ~1000
candidates.

Now somebody did log, sqrt, P99/p95 truncation and +several other
fancy transformations (one is from WOE, which I don't like). Each one
preserves original monoto. But how do we know one is better than
another? What procedures to run to compare them?
Thanks.
Paula