I am modelling volatility of international tourists arrivals from several source markets. I use mainly two methods ARIMA-GARCH or ARIMA-GJR models and SARIMA-GARCH or SARIMA-GJR models. Initially the estimates suggest that error terms of some models do not exhibit a normal distribution even though in the estimation I assumed it is normally distributed. I obtained the Bollerslev-Wooldridge standard errors in this case as it is said to be better than ordinary se. As some of the models do not have normally-distributed error term, I re-estimated all the models assuming a student-t distribution as it is recommended when the error term is non-normal. However, Bollerslev-Wooldridge standard errors were not available when using student-t distribution (In Eviews 10 software). Instead Huber-White se is available. I am wondering whether this is better than Bollerslev-Wooldridge standard errors or produce approximately similar outcomes. Any advice is much appreciated !!!
Hello S.C Thushara
There is no magic formula for errors to be normal identically distributed, in fact with no model you will find that the errors give you white noise, what you must ensure is that they do not have autocorrelation presence that is done with the pormonteu test in the lags and make sure that the squared residuals do not have arch effects with the Ljung-Box test. I recommend using distributions that have skewness and kurtosis to model the innovations of conditional volatility models, for that you must migrate from eviews to R or matlab.
Regards,
I saw in the book of chris brooks (2014) that adding a dummy variable (dummy=1 for the extreme year/data point) for extreme values of the residuals. This is when plotting the residuals of the regression. It can solve the problem. But the number of times you can add a dummy is limited. so it can also not lead to a better model with normally distributed errors
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Fixatives