Dear all
I have estimated a GARCH(p,q) model with multiple explanatory variables. Also, in order to solve problems regarding auto-correlation, partial auto-correlation and heteroskedasticity I have joined some ARMA and SARMA terms. However, my variance equation coefficients sum up to a little bit more than one. I have read that in order to solve this problem, one can raise the GARCH order (like from GARCH(1,1) to GARCH(2,2)) or even to add a @trend into the variance regressors. I have tried these procedures but i still have values greater than one.
Is these a real problem or can I ignore it? Are my results somehow skewed or biased if this occurs? How can I solve it?
I thank you all for the help in advance.
Yours sincerely, Afonso Rodrigues
Dear Pedro
I expecte that you get good results and you can do some risiduals testing (tests of good estimation) as ARCH Garch effect(must be no ARCH effect) and correlated of residuals (must be not correlated) also the test of normality (the residuals are normal distribution) if you get good results that is mean your estimated model is good.
Good luck
Abbas Lafta Kneehr thank you for your answer. Yes, I've conducted all those tests, and all of them provided the desirable results.
So do you think I should not worry about the fact that my variance equation coefficients sum up to more than one?
Yes I think that. especially when the residuals have not ARCH effect.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling