PROC GLM or PROC REG to get 95%CIs for independent variables

[smunigala]

Hi all,

I am working on a national database looking at continuous outcome and many independent predictors like age, race group, income category, insurance status, etc. So my predictors are a mix of  continuous and categorical variables. I need to check which of these variables are significant precitors and also get the 95% CIs for these predictors.

I used Proc GLM to test the association but I am finding difficulty in getting the 95% CIs for the independent predictors. I would appreciate if anyone can help me with this.

Also is there a way I can control the which category within a variables (eg. Race ) will be a reference category just like proc logistic?

ods rtf;
proc sort data = ulcer; by descending ulcernew descending agecat1 descending Racecat  descending INCCAT1 
descending INSURE descending MARRIED1 descending smoke1 descending backpain4 descending diabetes; run;
* Multiple linear regression;
proc glm data = ulcer order=data;
  class ulcernew agecat1 Racecat INCCAT1 INSURE MARRIED1 smoke1 backpain4 diabetes;
  model pcs36v2 mcs36v2 pf36 mh36 re36 rp36 sf36a vt36 
      = agecat1  Racecat INCCAT1 INSURE MARRIED1 smoke1 backpain4 diabetes ulcernew / solution ss3;
run;
quit;
ods rtf close;

Thanks,

Sat

[Rick_SAS · Accepted solution]

Yes. In the CLASS statement use the REF= option in parentheses to define the reference level:

class  Racecat(ref="white") MARRIED1(ref="yes") ...;

As for the confidence intervals, I assume that you want 95% CIs for the parameter estimates. Use the CLPARM option on the MODEL statement:

model pcs36v2 ... = agecat1  ... / solution ss3 CLPARM;

[smunigala]

Hi Rick,

Thanks for the reply. I am getting the 95% CIs for the parameters, but when I assign the ref=option, I think it is not working. Ref=option should be highlighted in blue right? It is not in my code. Do I have any error? I use SAS 9.4.

Thanks,

Sat

proc glm data = ulcer;
  class ulcernew agecat1 Racecat(ref="Black") INCCAT1 INSURE MARRIED1 (ref="1") smoke1 backpain4 diabetes;
  model pcs36v2 
      = agecat1  Racecat INCCAT1 INSURE MARRIED1 smoke1 backpain4 diabetes ulcernew / solution ss3 CLPARM;
run;
quit;

[Rick_SAS]

Don't worry about the color. It works. Run it and you'll see.

FYI, suboptions do not always get coloration in the text editor. For example the following uses the WHERE= data set option, which is a valid option but is not colored blue in my version of SAS Windowing Environment (DMS), although the WHERE= (and REF=) statements are colored blue in SAS Studio and in this Support Communities window:

proc glm data=sashelp.class(where=(weight<200));
class sex(ref="M");
model weight = height | sex / solution;
run;

[smunigala]

Thank you so much!