Dear all,
I want to know that when we run any GARCH model we firstly estimate a mean model and do its residual diagnostics to see arch effect. if arch effect found then run GARCH models.
mean model like with constant only, with AR or with MA or with ARMA.
Suppose i run a VAR bivariate BEKK GARCH model and get results in which mean model coefficients are insignificant but variance model coefficients are significant or insignificant. then in such case whether we should ignore mean model insignificant coefficients because we are checking for variance model and we have mean with residuals of mean model showing arch effect.
means if arch effect was there in residuals of mean model(Whether estimated through constant, AR or MA) we tested for garch model.
So now can ignore insignificant coefficients shown by mean model in GARCH model results.
thanks and regards..
This paper might help you:
Practical Issues in the Analysis of Univariate GARCH Models
http://faculty.washington.edu/ezivot/research/practicalgarchfinal.pdf
Yes. If you model return equation as e.g. ARMA(1,1) and find that e.g. MA coefficient is insignificant, then you shold simplify return equation, and in this case model it as AR(1).
Yes, the parameters of the mean equation have to be significant. Again, both mean and variance equations of the model must be adequate.
ARCH and GARCH coefficients have to positive and the stationarity condition should be hold.
Teena I also posted a presentation how to model ARCH/GARCH models! You can consult to it!
- Badges
- Science topic
- Similar topics
- Correction