Categorization of continuous variables

JovanaUniCredit — Thu, 03 Jun 2021 08:25:36 GMT

Hi community!

I have to do a categorization for continuous factors (e.g. using deciles, or different intervals consistent with the sample size) in order to provide more stability to the analysis of their relation with the target variable (which is not binary, but interval one).

If I decide to use deciles, each bucket should be composed by at least 30 observation in order to perform following discriminatory power test.

Is it possible to be done in Enterprise Miner?

I saw that there is Interactive Binning node, where interval target variable is converted to binary one. Is it correct to use this node, or there is other node more appropriate for this task?

In addition to this, if it is not possible to be done in Enterprise Miner, how it can be done in SAS EG?

P.S. If this is not correct forum to ask this question, please tell me which one is in order to get answer.

Thank you, Jovana

Re: Categorization of continuous variables

gcjfernandez — Fri, 04 Jun 2021 05:55:54 GMT

The following SAS Resources can be helpful:

Optimal Binning in SAS enterprise miner:

https://communities.sas.com/t5/SAS-Data-Mining-and-Machine/Optimal-Binning-in-the-Enterprise-Miner-Transform-Variables-node/td-p/213153

https://support.sas.com/resources/papers/proceedings15/SAS1965-2015.pdf

Interactive grouping in SAS Enterprise miner (Credit scoring license)

https://go.documentation.sas.com/doc/en/emref/14.3/p1qzwz7onopjqcn11uc04i18urg7.htm

Using HPBINN proc

https://blogs.sas.com/content/iml/2019/08/05/proc-hpbin-bin-variables-sas.html

The ideal method of binning depends on your analytical objectives, the properties of input variables and the target variable

topic Re: Categorization of continuous variables in Statistical Procedures

Categorization of continuous variables

Re: Categorization of continuous variables