partial F-test in PROC PANEL

[EmanuelS]

Hello Community,

I'm trying to perform a partial F-test. My question is: Is a specific group of variables together significant?

As an example: Test for the variables jan-dez (dummy-variables for each month):

y= ß1*did+ß2*treatment+ß3*secyear+ß4*jan+ß5*feb+...+ß15*dez

F-test:

ß4=ß5=...=ß15=0

I can do that in PROC REG with TEST, but in PROC PANEL it does not work.

Hope someone can help me.

Kind regards,

Emanuel

[SteveDenham]

Have you tried:

Test: jan=0, feb=0,mar=0,..., dez=0;

It appears that the syntax for PROC PANEL uses the variable names rather than the beta's associated with the variables.

Steve Denham

[EmanuelS]

Hi,

I tested that today, but I get an error: "the test ... cannot be implemented due to collinearity problems"

jan is not in the sample!

Emanuel

[ets_kps]

This is likely because Jan is not in the model as it was dropped due to the "dummy variable trap."  Make sure your tests refer to parameters that were actually estimated. -Ken