Absorb Fixed Effects and Cluster Standard Errors

econjoe — Fri, 18 Sep 2015 18:32:10 GMT

Is there a SAS equivalent to Stata's areg command? Specifically, I'm looking for a procedure that will replicate the following Stata command:

areg depvar indvar, absorb(id1) cluster(id2)

In this case id1 is nested within id2.

The proc genmod below clusters the standard errors at the id2 level, but is not able to absorb id1.

proc genmod;

class id2;

model depvar = indvar;

repeated subject=id2 / type=ind;

run;

The proc glm below absorbs id1 but cannot cluster the standard errors at the id2 level.

proc glm;

absorb id1;

model depvar = indvar / solution noint;

run;

Is there a simple way to both absorb id1 and cluster on id2?

Re: Absorb Fixed Effects and Cluster Standard Errors

SteveDenham — Fri, 25 Sep 2015 17:48:40 GMT

I am not real sure, but would some sort of hierarchical model, such as GLIMMIX work for what you want?

proc glimmix data=have;

class id1 id2;

model depvar=indvar;

random intercept/subject=id1;

random _residual_/subject=id2(id1) type=vc;

run;

Steve Denham