EViews.com

Hi, everyone, as everybody starts their questions here: econometrics and Eviews are very new for me! I want to estimate calendar effects, in particular day of the week effect and month of the year effect on one particular stock's return using Garch (1,1) model with the help of Eviews.
1. I have estimated stock return (for nearly 20 years) and stuctured dummy variables for each day of the week and each month of the year in eviews.
2. I've found out how to technically estimate it in eviews, but not sure how to write equation, I don't want to include constant into equation, but as I understood it will be included automatically. Is it right?
3. If it is the case,should I write the mean equation in the foolowing way: return c tuesday wednesday thursday friday or I may write return monday tuesday wednesday thursday friday?
4. should I do any kind of assuption test before running garch (1,1) regression? If yes, which?
5. What else I might have forgotten?

Please, do not laugh at me as I really need you help!!!
I would really appreciate your advices!!!

Dear ronni777

I am also not an expert in Eviews or Econometrics but I think I can give some correct ideas.

I think that you can't choose whether to add the constant term or not. In fact, when estimating a GARCH model you must make sure all your variables are stationary, so you should perform unit root tests like ADF, PP and KPSS. When performing those tests you will be able to see if the constant and trend are significant and, therefore, decide whether to include them or not.

As for the dummies I think the order you write them into the equation is not relevant. However, you must not forget that, if you are working with 5 week days, you can only include 4 day-of-the-week dummies. Regarding the month of the year, you can only use 11 dummies. there isn't a right day of the week or month to exclude, you will just have to choose a reason and stick with it.

Although is very likely that your data contains ARCH effects, and thus making the GARCH estimation adequate, you can firstly test for the ARCH effects so you can empirically prove the need to perform the GARCH model.

You may also need to go throw an ARIMA process to check for the need to include AR and\or MA factors.

When the GARCH model is estimated, you should run some diagnostic test, namely residuals diagnostics like ARCH LM test, Correlogram Square Residuals and Correlogram Q-statistcs.

I hope this can somehow help you.

Yours sincerely,
Afonso Rodrigues