Dear sir or madam:
I was wondering if these results show that my ardl model is stable. I mean if we see the cusum square graph the blue is outside the band and then comeback within the band.
Yours,
Joel
Dear Nip
Looking at the first graph there is suerly some instability in the model at least at 5% significance level. What I would recommend it to use Chow test for double-check.
best
Piotr
Chow test is implemented in most econometric packages I know so you dont need to do it by yourself. In case you want you should add not only dummy variable (it will represents the shift in constant) but also interactive variables (dummy*original variable)for all variables you have in the model. Than you have to run Wald test with restrictions for all new variables that all of them are jointly not statistically significant.
Dear Prof Cizkowicz
I'm using eviews and i ran a Chow test for independent variable, Y, using the regression output obtained from the command "LS Y C", where C is a constant.
Should i use the breaks suggested by the criterias by creating 3 dummy variables and then use them as fixed regressors in my ARDL model?
Yours.
Joel
CUSUM AND CUMSUM tests are based on recursive residuals of a regression. Respectively theey test the constancy of regressiom coefficients and variamce. The Chow test has a null that the model does not change starting at a predetermined observation. This observation usually coresponds to some policy change.
The tests are not applied to the dependent variables in a regression. Indeed one needs a range of values of independent variiables to estimate coefficients.
Are you interestwd in determining if some observations have excessive effect on coecficient estimafes? There are influence statististics that you might find useful.
- Badges
- Science topic
- Similar topics
- Mathematics