topic Proc rank decile binning in Statistical Procedures

Proc rank decile binning

KafeelBasha — Fri, 16 Jun 2017 09:41:33 GMT

Hello

In logistic regression after deciling using proc rank, I ran proc freq which returned below output

I am getting different frequencies in each decile, is it necessary that we should get equal number of frequencies in each decile?.

Thanks

Re: Proc rank decile binning

PeterClemmensen — Fri, 16 Jun 2017 09:49:48 GMT

Necessary for what? Use the GROUPS= option in PROC RANK if you want to have equal frecuencies in your groups

Small example

proc rank data=sashelp.cars out=carsRanked groups=10;
	var invoice;
	ranks group;
run;

proc freq data = carsRanked;
	tables group / nocum;
run;

Re: Proc rank decile binning

Rick_SAS — Fri, 16 Jun 2017 09:54:14 GMT

You will not necessarily get identical numbers in each bin. See the example (and cautionary tale) in the article "Binning data by quantiles? Beware of rounded data!"

Re: Proc rank decile binning

KafeelBasha — Fri, 16 Jun 2017 10:41:39 GMT

Thank you very much.

Lets say I want to calculate expected defaults(Y=0), I have 8417/16880=0.499 zeros, then can I say that the expected defaults is 2044*0.499=1019.956 for the decile 4?.

Thank you.