Garch "day of the week effect"

[jnkfoune]

Hello,

I am really new with Eviews and I would need some help to run a Garch Model on market daily return series to study the Day of the Week effect (volatility is or not higher on Mondays).

I am very in the rush and can't really spend much time on Eviews at the moment even though it look like very powerful and helpful tool.

Well, my aim is to use a GARCH model with 5 dummy variables (one variable for each trading day in the week) but I don't know yet how to set a Garch model and even less how to add dummy variables.

Thank you for your help, I stay free to add any detail

(I am using Eviews 7)

[trubador]

Assuming you have already structured your workfile as "daily - five day week" and the dependent variable as return series, one of the most convenient ways of estimating your model in EViews would be running the following code from the command window:

Code: Select all

arch(1,1) r c @ @expand(@weekday,@droplast)

where p=1 and q=1, and the constant term in the variance equation represents the Friday effect (in order to avoid dummy variable trap). Monday effect will correspond to the significance of the coefficent of (@weekday=1) variable.

[jnkfoune]

Thank you for your help, let me just ask you for one more detail, to make it clear for me :)

in my workfile, in order to structure the week days part, I should get 5 columns filled by zeros and put 1 in the correct colum (corresponding with the right day) for each line? (I mean that if the return is on a Monday, put 1 in the column 1 and 0 in the others)

Or it should work using 1 variable weekday that can take values from 1 to 5?

I Did not understand well the dummy variable trap...

Thanks again

[trubador]

@expand creates automatic dummy variables for use in equations without creating series in the workfile explicitly. In your case, @expand(@weekday,@droplast) creates the dummies just as you have described. It drops the Friday dummy (@weekday=5) in order to avoid perfect collinearity.

[jnkfoune]

I had just tried when you answered (I tried both "techniques" and realized on my own what did @expand) I guess that @droplast drops the Friday (which becomes the C variable right?) I will continue my research on dummy variable trap the the drop of the Friday as I still did not understand perfectly :D

Thank you again for your help! :)

[sandunsilva01]

Hi,
I'm a new one to this forum. I also doing a same research to examine the day of the week effect on ASPI index (Sri lanka). As you guys discussed above i have created a dummy variable for the week days using @expand command. But when i open my return series and that group dummy variable as a group and try to build up a GARCH(1,1) model, the following error message appears.
"ARCH estimation requires a continuous sample"
Why is this happening? Please help me....