The RATS Software Forum

Hi,

I hope this finds you well,

I would like to ask you about the test of multiple changes in linear regression model with lagged dependant variable.

when I apply the test of Bai& Perron, I find extreme different results for the number of structural breaks; BIC = 6 and LWZ = 0. what should I do in this case?

I don't understand the rest of the results with the regression on DZ(1,1)?

and to which critical values should I compare the results of the Shifting Regressors?

Thank you in advance,

Ahmed Sahloul

LWZ has a correction for degrees of freedom, which BIC doesn't. BIC for any least-squares model actually has a global minimum where you have enough parameters for a perfect fit and will generally get small just before that. I'm guessing that you have 6 x regressors nearly equal to the number of data points. If that's true, 6 is way too many breaks to consider. (It's probably way too many breaks in almost any model).

The D's are the dummied out regressors.

Hi,

Thank you for your reply,

I think that the test may not be the best guide for me to determine the number of breaks.

I tried to follow the economic development of the economy under study, adding a dummy variable for each variable "in each equation" to correct the parameters instability, when necessary. Therefore, the dummy variables correct for different dates.

What do you think about this approach?

Thank you and best regards