EViews.com

Dear all,
I have just estimated an AR markov-switching model. I get two coefficients for my regime-dependent variable of interest (regressor) which is a dummy.

I would like to test whether such coefficients are equal across the two regimes instructed: C(2)*ID_AUCB1=C(4)*ID_AUCB1, where c(2) is in the first equation (regime 1) and c(4) is in the second (regime 2).

When I try to do this through "Coefficient Diagnostics>Wald Test-Coefficient Restrictions", after I write "C(2)*ID_AUCB1=C(4)*ID_AUCB1" in the dialog box, it pops out a second dialog box "Observation" where I have to specify Date or ID.

1) I do not understand this, it seems as I can test it only for a specific date. Am I right?

2) I want to test it for the whole sample used for the estimation, is there a way to do it?

3) Am I missing something? Is it possible to test linear restrictions across regimes?

I did not find any reference in the manual.

Thanks for helping about this.

Gianluca

Just enter

c(2)=c(4)

which will test the coefficient equality.

You were trying to test whether the coefficient times the series were the same in the two regimes. But since that restriction differs by observation, we prompt to ask you the date ID at which you wish to test.