topic Re: regression code for categorical predictors and regresee(predictee?) in SAS Programming

regression code for categorical predictors and regresee(predictee?)

JKCho — Wed, 10 Jul 2019 20:59:24 GMT

Hello!

I need to use one model used in previous papers but, of course, do not know the code for that,

It is the regression of credit ratings on some indicator variables(like loss or income) and others(like sales).

As there are more than 20 different credit rating scores and I need to use some indicator variables, I am lost how to make the code.

I find many about using categorical predictors but unsure about categorial regressees. Maybe using Logit/Probit thing...

e.g. : Credit Rating = b0+b1Sales+b2Loss indicator+b3dividend indicator...

If you already have worked on credit rating regression like this, please share your wisdome!

Thank you!!!

Jerry

Re: regression code for categorical predictors and regresee(predictee?)

PaigeMiller — Wed, 10 Jul 2019 21:41:02 GMT

I would use Partial Least Squares regression instead of ordinary least squares regression, and your indicator variables go into the CLASS statement, all variables go into the MODEL statement.

Documentation with examples:

https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_pls_examples01.htm&docsetVersion=15.1&locale=en

Logit and probit are only used when the Y-variable is categorical; it is not used when an X-variable is categorical and the Y-variable is continuous.

Re: regression code for categorical predictors and regresee(predictee?)

Ksharp — Thu, 11 Jul 2019 12:28:15 GMT

Buy Naeem Siddiqi 's book

Intelligent Credit Scoring: Building and Implementing Credit Risk Scorecards, Second Edition

https://www.sas.com/store/books/categories/business-concepts-leadership/intelligent-credit-scoring-building-and-implementing-credit-risk-scorecards-second-edition/prodBK_69286_en.html

Re: regression code for categorical predictors and regresee(predictee?)

JKCho — Thu, 11 Jul 2019 14:19:02 GMT

Thank you for recommending the book!

Re: regression code for categorical predictors and regresee(predictee?)

JKCho — Mon, 15 Jul 2019 15:46:04 GMT

Thank you PaigeMiller!
I used the PROC PFS and notice that it does not work with the fixed variable code like absorb identifier;

Because my model has some control variables too I want to put them in the model. Do you happen to know how to use control variables within PLS?

Re: regression code for categorical predictors and regresee(predictee?)

PaigeMiller — Mon, 15 Jul 2019 15:55:01 GMT

Control variables ought to be included in the model the same as any other variables.

Re: regression code for categorical predictors and regresee(predictee?)

JKCho — Mon, 15 Jul 2019 16:19:48 GMT

Last question!
how canan I also put fixed effect code such as absorb ; with PROC PLS? I have known that the ABSORB statement does not work with PROC PANEL and PROC PLS. may need sth else?

Re: regression code for categorical predictors and regresee(predictee?)

PaigeMiller — Mon, 15 Jul 2019 16:47:58 GMT

There is no ABSORB statement in PROC PLS.

If you really have a class variable that needs an ABSORB statement in PROC GLM, you'd have to include it in the model as a class variable in PROC PLS.