topic Re: Logistics Regression with nominal inputs of multiple levels in SAS Data Science

Logistics Regression with nominal inputs of multiple levels

MLAC — Fri, 15 Nov 2019 20:38:33 GMT

I am working on a binary logistic regression problem with nominal attributes that have multiple levels, for example:

Age:

young(10-18)

adult(19-45)

mid_age(46-65)

senior(65+)

How SAS handle a nominal attribute that has more than 2 levels in Logistic Regression?

Re: Logistics Regression with nominal inputs of multiple levels

PaigeMiller — Fri, 15 Nov 2019 20:53:02 GMT

Under the hood, it will turn these four different age groups into dummy variables and then fits the model. In your case, where there are four different age groups, it would use three dummy variables.

Re: Logistics Regression with nominal inputs of multiple levels

Reeza — Fri, 15 Nov 2019 20:53:29 GMT

Depends on what you specify as your parameterization method. There are several methods that are noted in the documentation, each are handled slightly differently, the most common being dummy coding 0/1 for all variables and dropping one variable.
https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_logistic_syntax05.htm&docsetVersion=15.1&locale=en#statug.logistic.classstmtunbal