12-31-2014 12:14 PM
I want to be able to discretize (bin) 1000's of continous (numeric)-level variables in a wholesale fashion and create new SAS nominal-level variables for each. Does base SAS or SAS/Stat (not EMiner) have a binning Proc that is similar to the R package 'discretization' (attached)? I need something that creates optimal bins given some criterion for "optimal"; I don't just want equal spaced or equal population bins. And it would be nice to have a WOE metric also produced for each bin. Thanks
01-01-2015 01:36 AM
I do not think there is equivalent proc in SAS.
But some of these functionality you may find other have coded or used another method for that purpose.
So if you know exactly what you want and for what you may get more help in the community.
Also, it may deserve to take a look at Having an EFFECT: More General Linear Modeling and Analysis with the New EFFECT Statement in SAS/STA...
01-01-2015 03:19 AM
There are SAS papers on that question of continuous/categorical like: http://support.sas.com/resources/papers/proceedings09/248-2009.pdf
There is an other approach with SAS that could do the job of recoding. That is using SAS formats when the proc is supporting the formatted value and not the real one.
This concept is missing in many other languages like R.