SAS PROC REG REFERENCE GROUP

[AlainX]

Dear All,

Consider a CLASS variable, GENDER, with values F and M. I understand that in SAS, by default, these values are arranged in ascending alphanumeric order which results in M being the last level, and therefore the reference level. Suppose alternatively that we want to set F as the reference level.

Below, I list the coefficients for each case.

I understand that the 2 cases are "equivalent".

However, since in each case the estimated equations are not the same:

y=20.1903-1.6454Gender-0.2917Height

y=18.5448+1.6454Gender-0.2917Height

they provide different probabilities, if they will be manually applied for a new observation.

I apologise in advance for the naive question.

A.

[PaigeMiller]

This cannot be the output from PROC REG. What PROC are you actually using? Show us the code used.

[AlainX]

Hi. I am using SAS Studio On Demand. I found the example here: https://support.sas.com/kb/37/108.html

The reason why I am asking this question, is the following one:

https://communities.sas.com/t5/Statistical-Procedures/Interpretation-of-Logistic-Regression-Coeffici...

Thanks in advance.

[PaigeMiller · Accepted solution]

This is not PROC REG. It is (perhaps) the Regression function in SAS Studio.

The models are identical. But you have written them wrong.

y=20.1903-1.6454*(Gender='F')-0.2917*Height

y=18.5448+1.6454*(Gender='M')-0.2917*Height

For females, the first model equates to y = 20.1903-1.6454-0.2917*Height = 18.5449 - 0.2917*Height

For females, the second model equates to y = 18.5448 - 0.2917*Height

These are identical (except for roundoff error).

[AlainX]

Dear PaigeMiller, 100% clear now. Millions of thanks.