Dear Sir or Madam,
Is it correct to considerer structural break in the data by adding a Dummy2000 (which adquire values of 1 since the year 2000) and an interaction term, Dummy2000*X1.
Y = intercept + D2000 + X1 + D2000* X1.
Can i add these dummies and interaction terms as fixed regressors in a ardl model?
Should I delete them from the equation if they are not significant?
Your opinion will be highly appreciated.
Yes. you should consider the structural break in to your model by some method such as dummy variable. Of course before this you should test structural break by CUMSUM test or break even point or chow test
Estimado Nip Nip.
Usted puede utilizar variables dummy para medir la estabilidad estructural del modelo seleccionado, debe establecer el período que será igual 0 y el período igual a 1.
Yes. you should consider the structural break in to your model by some method such as dummy variable. Of course before this you should test structural break by CUMSUM test or break even point or chow test
Let's say i found a break in year 2000, should i add it conserve it even if is not significant in the ARDL model?
What about the slopes. I have to independent variables, Should i add two interaction terms, i.e., D2000*X1 and D2000*X2 and conserve them even if they are not significant?
there is no need of interaction term, however add dummy of structural break in model.
the more appropriate way is to visualize the dependent variable, if there is possible of structural break then add it into the model.
Dear Nip, you should always be concerned about the structural breaks. The type of dummy you presented captures, if statistically cannot be rejected, a break in the intercept term. From time to time, the structural break could also affect the slope and a new data generating process or a new regime could have been started. About the abovementioned question, instead of interaction terms, if sample size is enough, estimate two sub-samples seperately.
Best.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series