hi all,
i need help to find out if my series include seasonal effect. I am trying to test with dummy variables? As i learned before in monthly data i have to use 11 dummies and estimate this equation
y d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11
Here are my questions ;
- is constant should take a part in this equation? like (y c d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11)
- what the f values have to be if there is seasonality?
Thanks and b. regards
seasonality test
Moderators: EViews Gareth, EViews Jason, EViews Steve, EViews Moderator
Re: seasonality test
The following code may be more useful, since it uses seasonal dummies specifically designed for this purpose.
You can use the whole set of dummies if you do not include a constant. However, if you decide to include the constant, then you should drop one of the seasonal dummies.
Code: Select all
y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12)
You can use the whole set of dummies if you do not include a constant. However, if you decide to include the constant, then you should drop one of the seasonal dummies.
Re: seasonality test
thanks for answering, i got the difference you mentioned. but still i dont know what the f values of dummy variables should be if there is seasonality.
Also is there any easier way to test seasonality? i tried census x12 but couldn't understand the whole output. it says something like :
Test for the presence of seasonality assuming stability.
Sum of Dgrs.of Mean
Squares Freedom Square F-Value
Between months 1733.2908 11 157.57189 4.268**
Residual 3987.3116 108 36.91955
Total 5720.6023 119
**Seasonality present at the 0.1 per cent level.
is that table where i should look at?
Also is there any easier way to test seasonality? i tried census x12 but couldn't understand the whole output. it says something like :
Test for the presence of seasonality assuming stability.
Sum of Dgrs.of Mean
Squares Freedom Square F-Value
Between months 1733.2908 11 157.57189 4.268**
Residual 3987.3116 108 36.91955
Total 5720.6023 119
**Seasonality present at the 0.1 per cent level.
is that table where i should look at?
Re: seasonality test
Census X12 method is a more formal procedure to handle the seasonality. If seasonality is the primary focus of your study, then I suggest you to continue with X12 method. The output includes results of many diagnostic tests and analyses that pertain to seasonality issue, so it generates a pretty long document. Yet, it is fairly self-explanatory. In your case, for instance, seasonality is detected and is significant at .001 level. When you scroll down the output, you'll also see the result of the same test after the seasonality is removed. If there is no indication of residual seasonality, then you can safely use the seasonally adjusted series.
Re: seasonality test
i compared the results (regression with dummies & x12) and still cant find out ...
i tried those ones:
- y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12) this one comes with 0.000 prbability value for all dummies
- y c @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) and this one all comes with values >
0.05 except for constant's
why those two estimation gives different results? which one is true? Also as i said the x12 method says there is seasonality.
thanks a lot
i tried those ones:
- y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12) this one comes with 0.000 prbability value for all dummies
- y c @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) and this one all comes with values >
0.05 except for constant's
why those two estimation gives different results? which one is true? Also as i said the x12 method says there is seasonality.
thanks a lot
Re: seasonality test
i tried this situation on different series and i think the correct estimation to test seasonality with dummies should be :
y c @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11)
i am waiting for any other suggestions..
thx
y c @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11)
i am waiting for any other suggestions..
thx
Re: seasonality test
You can think of the constant as the coefficient of a series full of ones. Therefore, these two specifications naturally produce different output. If you do not include the constant, each coefficient will give you the marginal effect of each month. However, interpretation of the results changes when you include the constant. This time, each coefficient shows the relative effect of each month to that of reference month (i.e. missing/dropped category). It really does not matter that much which specification you use as long as you are aware of this difference. As I said before, if seasonality is your primary goal, you should better use X12 method.
Re: seasonality test
i just want to know if there is seasonality and then, which month(s) cause that. i want to analyse that with dummies. Thats all ..
In that case as i understand from your explanation i have to estimate without constant. The x12 method says there is seasonality but when i estimate this:
y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12) all the p values comes with 0.000
please tell me what should i comprehend from these probability values ?
here i attached the file incase it helps
THX
In that case as i understand from your explanation i have to estimate without constant. The x12 method says there is seasonality but when i estimate this:
y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12) all the p values comes with 0.000
please tell me what should i comprehend from these probability values ?
here i attached the file incase it helps
THX
- Attachments
-
- untitled.WF1
- (77.11 KiB) Downloaded 647 times
Re: seasonality test
As you should know, p-values smaller than a given confidence level (say 0.05) denote the significance of the associated parameter. Since your series is not stationary, it might be a good idea to add a trend variable (i.e. @trend) into your equation besides seasonal dummies. Also in that case, seasonality seems to be a strong component. When you estimate the seasonality this way, you assume that the seasonal components are additive and do not change over time. However, most macroeconomic time series display a multiplicative and a moving seasonal pattern, both of which are handled effectively through the X12 procedure. You can obtain the seasonal effect of each month via clicking the "Final seasonal factors ( _SF)" as the "Component Series to Save" in the X12 options window.
Re: seasonality test
why everytime i use this y @seas(1) @seas(2) @seas(3) @seas(4) @seas(5) @seas(6) @seas(7) @seas(8) @seas(9) @seas(10) @seas(11) @seas(12) i get syntax error ?
Who is online
Users browsing this forum: No registered users and 26 guests