Collinearity

[sguerra]

Hi users,

Here's my question.

Currently I'm working detecting and correcting parameter instability in linear models using APBREAKTEST.src.

One of the standard methods for correcting this problem is to use dummy variables.

Suposse that the model Yt = c + Xt + Zt + Zt{1} + Wt, have some parametter instability. Lets asume that the problem is presented in all parametters.

We create a dummy variable using the date sugested by APBREAKTEST.scr, then create more dummies that interact the first dummy with all the regressors, and then, if we test again including all the new variables, the problem should be solved.

One problem that may arise is that the dummies concerning with the variable Xt and its lag could have some collinearity.

My problem is that I have around 4 lags of Zt variable and the collinearity problem is in fact causing me some troubles.

I've seen that some packages such as STATA, when detect Imperfect-Collinearity, drop at least one of such variables.

Does anyone knows how to replicate that kind of solution (detecting and dropping) in RATS?

Thanks in advance








Some of these dummy variables could show some collinearity problems. I have seen that some packages such as STATA

[TomDoan]

RATS doesn't drop the redundant variable(s) from the regression, but they end up in the output with zero coefficients and zero standard errors, and the degrees of freedom are adjusted so as not to count the extra variables.