Page 1 of 1
seasonal analysis
Posted: Mon Sep 07, 2009 6:28 am
by tomasi
Hi all,
unfortunately I haven't dealed with EViews so far and also my econometric knowledge isn't the best. Therefore I would like to take use of this forum.
I would like to do an analysis of gasoline prices and therefore test whether there are any seasonalities or not. I have daily data of petrol prices available and would like to see if e.g. the prices are on monday lower than on friday!
Could anybody give me instructions or hints how to do that?
Many thanks in advance!
Re: seasonal analysis
Posted: Mon Sep 07, 2009 7:06 am
by trubador
More formal seasonal adjustment procedures are only available for monthly and quarterly data. You can construct dummies for each day of the week to conduct a simple test for seasonality. If your workfile is structured as "Dated-regular frequency/Daily 5-day week", then you can enter the following into your equation dialog:
Code: Select all
your_series (@weekday=1) (@weekday=2) (@weekday=3) (@weekday=4) (@weekday=5)
Explanatory variables correspond to dummies for Monday, Tuesday, Wednesday, Thursday and Friday, respectively.
Re: seasonal analysis
Posted: Mon Sep 07, 2009 7:19 am
by tomasi
Thank you very much for your fast response!
I have selected a "Dated-regular frequency/Daily 7-day week" and made a series of dummies for the single days, as monday=1, tuesday=2, etc.
Do you think it's a disadvantage to choose a 7-day week? What is the "@" for and why in brackets?
Is your suggestion the same as I've already done, like: my_series c weekday=2 weekday=3 weekday=4 weekday=5 weekday=6 weekday=7
or: my_series weekday=1 weekday=2 weekday=3 weekday=4 weekday=5 weekday=6 weekday=7
In addition I made a dummy for holidays with a "1" if there is a holiday and a "0" if not. Do you have an idea how to check that?
Many questions :(
Re: seasonal analysis
Posted: Mon Sep 07, 2009 7:41 am
by trubador
"@weekday" is a function of EViews that returns the day of the week and therefore is a quick way of generating dummy variables. Brackets are not necessary from the technical point of view. They are just used for the sake of exposition. Finally, if you use a "7-day week" structure and generate dummies for each day, then you do not have to generate an additional dummy for the weekend. Since the former specification already tests the effects of Saturday and Sunday, introducing an additional dummy for the weekend will create redundancy...
Re: seasonal analysis
Posted: Mon Sep 07, 2009 7:48 am
by EViews Gareth
It might be worth having a dummy variable for national holidays.
And as an alternative to the method Trubador stated, you can generate the dummies automatically for weekday by using the following equation specification:
Re: seasonal analysis
Posted: Wed Sep 09, 2009 4:21 am
by tomasi
Thank you very much for you replies and the good advices!
In addition to the seasonal analysis I would like to test for a relationship and asymmetries respectively between oil prices and gasoline prices. Has anybody an idea how to do that? Which possibilities do I have to create an equation in order to investigate the relationship between these series? How could the equation look like?
And which tests do I have to do at the beginning to get sure that my data is appropriate?
Thanks in advance!
Re: seasonal analysis
Posted: Wed Sep 09, 2009 5:29 am
by trubador
I think at this point you should better refer to econometrics and/or time series analysis textbooks since they provide more detailed information on such issues...