Help with Proc Surveyreg and Test Statement

[therock]

Hi Everyone,

I am trying to test the effect on performance of different publicly traded companies based on whether they have audit committee or not after the passage of Sarbens-Oxley Act (SOX) of 2001 based on if they are listed on New York Stock Exchange (NYSE), NASDAQ, American Stock Exchange (ASE), or other exchanges. I have panel data that starts from 1996-2012. So here is my code:

proc surveyreg data=problem;

cluster firm;

class year exchange;

model performance = Audit PostSox Exchange ControlVariables Audit*PostSox*Exchange / noint solution;

test Audit*PostSox*NYSE = Audit*PostSox*NASDAQ = Audit*PostSox*ASE = Audit*PostSox*Other;

run;

The test statement does not work. I need help with that please. Can somebody tell me what I am doing wrong and how to fix it?

If I use the BY statement in the program, then I need to test the same thing:

proc surveyreg data=problem;

cluster firm;

class year;

BY Exchange

model performance = Audit PostSox ControlVariables Audit*PostSox / noint solution;

test Audit*PostSox for NYSE = Audit*PostSox for NASDAQ = Audit*PostSox for ASE = Audit*PostSox for Other;

run;

Which model is better and accurate? Also how do I test the coeffecient in SAS?

Thanks,

the rock 

[ballardw]

As a bare minimum you appear to be referencing variables in the test statement that do not appear in the model: NYSE NASDAQ and ASE.

Second I don't believe that the = sign belongs where you are attempting to use it.

I am guessing that all you actually want on the test is Exchange.

[therock]

Thanks for reply! 

Because Exchange is a class variable, I want to test difference between NYSE, ASE, and NASDAQ. So in the model statement I just put Exchange, but in realty it will create a dummy variable for each exchange.

Sorry if the original statement was confusing. Hope this clears up!

Thanks!

[therock]

I forgot to add. I want to test the interaction term between all three variables.