PROC GLM: effects coding?

[Ruth]

Hi chaps,

I am running an analysis of multiple linear regression using PROC GLM, with the dataset mainly consisting of categorcial variables.

Do you know that if PROC GLM has the effects coding (-1, 0, 1) for the design matrix. As compared to the dummy coding (0, 1), the effects coding offers the advantage of comparing between each category and the grand mean.

Thanks in advance.

[Rick_SAS · Accepted solution]

For everything about effect coding in SAS, see http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_introcom_a00...

GLM doesn't have the coding you want, but GENMOD does, and can do GLM-type models.

Or you can use the OUTDESIGNONLY OUTDESIGN= option in PROC LOGISTIC to output dummy variables for the explanatory variables, and this use those dummy variables in GLM. (Use anything for the response variable, like a constant vector.)

Rick

[Ruth]

Thanks