topic Create dummy variable for group in Statistical Procedures

Create dummy variable for group

Pt_Prof — Mon, 05 Oct 2015 18:58:59 GMT

Hello

I am looking at the interaction between age (dichotomized as above OR below 65 years) and sex (men versus women) on the functional outcomes of therapy. In my model, I would like to insert 3 independent variables: age sex agexsex

While creating the interaction variable, I just created a column of data representing agexsex. Following this, I conducted regression analysis using SAS code:

Proc reg data = Knee_gender_interaction2;
model IADL_6months = age sex agexsex/stb clb;
model functions_6months = age sex agexsex/stb clb;

run;

The problem I run into is that SAS is not allowing me to get the beta estimate for the interaction term agexsex, by suggesting that the age-sex is the linear product of age and sex which are also present in the model.

I read somewhere that I should create a summy variables to test the effect of all 3 independent variables: age, sex, and agexsex.

Can somebody suggest teh SAS coding for creating the dummay variable and subsequent regression model?

Thank you much.

Saurabh

Re: Create dummy variable for group

Reeza — Mon, 05 Oct 2015 19:19:04 GMT

Have you tried proc GLM with CLASS variables?

Re: Create dummy variable for group

thwasser — Mon, 05 Oct 2015 19:54:13 GMT

Did you create your sex indicator as a {0 ,1 } indicator? If not, do so-- and use this new sex indicator to also create your interaction product. Then retry in proc reg.

Re: Create dummy variable for group

lvm — Mon, 05 Oct 2015 19:55:02 GMT

See the answer to a similar type of question posted here.

https://communities.sas.com/t5/SAS-Statistical-Procedures/How-to-test-whether-the-differences-between-two-regression/m-p/227927#U227927