Information Value calculation

#1
Hi guys, I am reading a paper regarding modeling fractional outcomes -

http://support.sas.com/resources/papers/proceedings14/1304-2014.pdf

and I encountered a misunderstanding. Particularly, on the pages 4 and 5, the authors create frequency tables based on percentiles and in those tables they demonstrate a statistics called "Information Value".

I know this statistics because it is commonly used in credit scoring modeling and the formula for calculation is following:

http://ucanalytics.com/blogs/information-value-and-weight-of-evidencebanking-case/

So, the calculation is based on "Good" and "Bad" cases in particular groups/bins. Thus, your depended variable must be 0 or 1.

Please, could anybody explain me how the authors calculated Information Value when their depended variable (Y) is continuous??

Do you have any suggestions how could they come up with the information value?

Thank you very much
 
Last edited: