PROC REG output a specific F-test

[jl2978]

Hi,

I'm doing a regression by groups. For each regression, I'm also doing an F-test. I encounter two problems with the F-test.

First, the F-test will be v3 * myvar + v4 =0, where myvar also varies for each group.

Second, how can I output the F-test to the table?

I'm trying to write the following code, but it didn't work... 

Thanks! 

proc reg noprint data=reg_in outest=reg_out ;
  	model dep = v1 v2 v3 v4;
  	by group_id myvar;
  	t1: test v3*myvar+v4 =0;
run;

[PaigeMiller]

The TEST statement has to include variables in the model. It cannot include variables in a BY statement.

Although you could re-parameterize the whole thing to get a model that is equivalent to having MYVAR in a BY statement, I still don't think you use the TEST statement in the way you want. I think you can test constant*v3+v4=0, but I don't think you can test something like variable1*variable2+variable3 = 0.

[PGStats]

You could get that test by rescaling v3:

data test_in_2;
set test_in;
v3_myVar = v3 / myVar;
run;

proc reg noprint data=reg_in_2 outest=reg_out ;
  	model dep = v1 v2 v3_myVar v4;
  	by group_id myvar;
  	t1: test v3_myvar + v4 =0;
	ods output ParameterEstimates=PE TestANOVA=TA;
run;

Datasets PE and TA will contain the estimates and test results and also the value of myvar so that you can recalculate the v3 statistics.

(untested)