Day of the Week Effect with Dummy Variables, Garch(1,1)

[Mar93]

Hi everyone, I want to do test for day of the week effect with a Garch(1,1) model and dummy variables for the weekdays (Monday to Friday). I need help with constructing the workfile. In my Excel file I have the daily returns and also a column with numbers 1 to 5 for the weekday and four columns with dummy variables for days 2-5. When I import my excel file into eviews every column becomes a variable. I don't get how I need to transform my workfile in order to apply garch (1,1). Thanks for any help :) Does someone have a completed day of the week effect analysis in eviews with Garch(1,1) that you could send to me and which would be an example for me? that would be great.

my excel file looks like that:

Date R Weekday D2 D3 D4 D5
01/04/2001 4 0 0 1 0
01/05/2001 5 0 0 0 1
01/08/2001 1 0 0 0 0

[EViews Gareth]

You don't need the dummies in Excel, they're easy to generate in EViews. But having them there won't hurt. You shouldn't need to transform your workfile at all.

[Mar93]

Thanks for your immediate reply. But I am a bit confused about how exactly my workfile needs to look like in order to be able to apply the Garch(1,1) model.
I mean if I just have the return series with 5-Days-week, and then generate the dummies in excel or eviews, how does eviews know which date is a Monday, Tuesday or one of the other days?

[EViews Gareth]

Any non-panel workfile will do.

[Mar93]

Ok. But I'm still a bit confused. I haven't used eviews before and don't know how to prepare my workfile for the garch model. so I just take my return series from excel (so the return on that day and the corresponding date) and import it in eviews. then I create the dummy variables for the 5 weekdays. And then I can estimate the Garch model? does eviews then already know which date is a Monday, Tuesday or any of the other days? or do I need to de something else before estimating the garch model? thanks

[EViews Gareth]

Assuming you opened that Excel file in EViews, yes EViews knows which days are Mondays etc... EViews has a deep understanding of dates.

[Mar93]

ok great :) I have done that now. Is it correct that I only need to have dummy variables for days 2 to 5? because of the dummy variable trap?
and then I go to "estimate equation", choose arch, and in the specification I just put "r d2 d3 d4 d5"? and that will generate my garch (1,1) model?

[EViews Gareth]

That's how you estimate GARCH, yes.

[Mar93]

Ok :) thanks so much for your help.
Could you have one last look at what I get when I estimated the garch model (I've attached a printscreen). is this the correct final output? (BM is my return series)
thank you so much

[EViews Gareth]

That is correct GARCH output, yes.

[Mar93]

That was so much easier then I expected. Thank you very much :)

[Mar93]

One last question: What happens with Monday (D1)? I've seen others also putting d1 in the equation ("r d1 d2 d3 d4 d5"). what's the difference? In another topic someone received this output (he has d1 in the output. only the left one is relevant for me (GArch 1,1)). how can I find out anything about d1 when I only get the above output where d1 is not included? and what about c? I saw someone use this formula: arch(1,1) r c @ @expand(@weekday,@droplast), which also includes c.

[EViews Gareth]

That is a question of an econometric nature

[Mar93]

yeah sorry that has nothing to do with eviews itself. I'm going to review my lecture notes about the constant etc.
But you would advise me to put only d2 d3 d4 and d5 in the equation and d1 not, is that correct?

[startz]

Either the constant and d2-d5, or all five dummies.