how to run the regression if the dependent variable is a ranking number

[MR_Xishuai]

Hi all,

I am having a question about how to run the regression if the dependent variable is a ranking number?

For example, I want to look at what kind of factors will determine a person's education level. So my dependent variable:

EDUCATION= 1: PhD

                      2: Master

                      3: Undergraduate

                      4: High School

                      5: less than high school.

My independent variables are some continuous variables or dummy variables let's say: the person's IQ, his parent's salary...

I know that if the dependent is a dummy variable which has only 1 or 0, I can use Logit/Probit model. And if the dependent variable is categorical and cannot be ordered in any meaningful way, then I could use multinomial logistic regression.

But here the dependent variable has more than 2 categories and it is a rank which mean less value is better (for example, 1 is better than 5).

So how could I run this regression? Thanks a lot for you help.

[M_Maldonado · Accepted solution]

Hi Xishuai,

SAS regression procs support ordinal targets. If you are using SAS Enterprise Miner, all you need to do is set the role of your variable EDUCATION to target, and the level to ordinal. Then add a Regresion node to perform ordinal logistic regression.

If you are not using Enterprise Miner, here a good resource for "ordinal logistc regression".

     link: SAS/STAT(R) 9.2 User's Guide, Example: Ordinal Logistic Regression

Good luck!

Miguel

[MR_Xishuai]

Hi Miguel,

Thanks a lot for your detailed answer and yes, the ordinal logistic regression is definitely what I want.

Cheers,

Xishuai