Proc surveyreg

astudent — Wed, 03 Jan 2024 20:46:07 GMT

Hello,

I am running both simple linear regression and multiple linear regression using proc surveyreg. I have 2 types of variables: Total score (continuous dependent variable) and Q (binary independent variable coded as 0=negative response and 1=affirmative response) the log looks alright, there is no error message or any special note. I was expecting to get regression coefficients table which I got but instead of one coefficient for each independent variable, I keep getting 2 coefficients (one for 0 and another for 1). I want SAS to use the negative responses as the reference group and give me the coefficients for the affirmative responses.

In addition, I want to do a backwards selection multiple linear regression. I have tried to add this specification in the code as it is usually done in the proc reg procedure but it is not working. If I could get ideas on how to do this procedure for complex data, I would greatly appreciate.

Here is a screenshot of the coefficients table I am getting for multiple linear regression:

This is the reference code from which I derived my multiple linear regression code:

PROC SURVEYREG data=analysis_data nomcar;
STRATA sdmvstra;
CLUSTER sdmvpsu;
WEIGHT wtmec4yr;
CLASS  sex ethn smoker dmdeduc bmicatf;
DOMAIN eligible;
MODEL  lbdhdl= sex ethn ridageyr smoker dmdeduc smoker bmicatf/solution;
run;  

Here is a screenshot of the simple linear regression results:

Here is a screenshot of the simple linear regression results:

Here is the reference code I used for the simple linear regression:

PROC SURVEYREG data=analysis_data nomcar;
STRATA sdmvstra;
CLUSTER sdmvpsu;
WEIGHT wtmec4yr;
CLASS   bmicatf;
DOMAIN eligible;
MODEL   lbdhdl= bmicatf/ solution;
run;

Re: Proc surveyreg

ballardw — Thu, 04 Jan 2024 01:51:56 GMT

Suggestion when asking questions about output: Reference your variable names not what the variable means to you. You do not have anything in the code that shows a Q. So we have absolutely no idea which would be affected.

You would specify the reference level in the Class statement. For a specific variable it would look like:

Class somevar (ref='0')

topic Proc surveyreg in Statistical Procedures

Proc surveyreg

Re: Proc surveyreg