topic PROC GLM documents about two-way fixed effects explanation in Statistical Procedures

PROC GLM documents about two-way fixed effects explanation

Phil_NZ — Thu, 18 Feb 2021 22:12:23 GMT

Hi everyone,

I am normally using STATA to run the regression, now after knowing the aesthetic of SAS, I changed to use the SAS EG.

Regarding running the regression with two-way fixed effects, we normally use PROC GLM as the sample below:

proc glm data=merge_sample;
    class type yr;
    model wtot_ass = fir_age wfirm_size pt type yr/solution ss3;
run;
quit;

But when I read the document from the link (https://support.sas.com/rnd/app/stat/procedures/glm.html), I did not see any guidance relating to two-way fixed effect.

And one more question, when I want to run the regression above without intercept, what should I adjust the code?

Warm regards.

Re: PROC GLM documents about two-way fixed effects explanation

PaigeMiller — Fri, 19 Feb 2021 12:17:43 GMT

@Phil_NZ wrote:

I am normally using STATA to run the regression, now after knowing the aesthetic of SAS, I changed to use the SAS EG.

Regarding running the regression with two-way fixed effects, we normally use PROC GLM as the sample below:
proc glm data=merge_sample;
    class type yr;
    model wtot_ass = fir_age wfirm_size pt type yr/solution ss3;
run;
quit;
But when I read the document from the link (https://support.sas.com/rnd/app/stat/procedures/glm.html), I did not see any guidance relating to two-way fixed effect.

Do you mean two-way interaction?

And one more question, when I want to run the regression above without intercept, what should I adjust the code?

Without strong justification based on the logic of the problem, running regressions without an intercept is a bad idea, and doesn't usually fit the data well. So, do you have such a strong justification?

A better link to PROC GLM documentation is: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=statug&docsetTarget=statug_glm_toc.htm&locale=en

Re: PROC GLM documents about two-way fixed effects explanation

STAT_Kathleen — Fri, 19 Feb 2021 14:49:43 GMT

If you are estimting two-way fixed model for cross section and time variable you can consider using either PROC PANEL in SAS/ETS or PROC GLM or PROC MIXED in SAS/STAT. The following SAS Usage Note

http://support.sas.com/kb/22851

demonstrates the syntax for estimating fixed and random effects models using PROC PANEL and PROC MIXED.

The NOINT option in the MODEL statement will suppress the intercept from the model. For example,

MODEL Y = x1 - x10 state year / noint solution;

The PROC MIXED procedure is documented in the ONLINE DOC under SAS/STAT User's Guide. You may access the PROC MIXED documention:

http://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_mixed_syntax.htm&docsetVersion=15.1&locale=en

and the go to SAS/STAT--SAS/STAT User's Guide-- The Mixed Procedure.