test of seasonnality

[Mick Kountras]

Hi everyone,

I would like to have a simple and effective method to check if a series is seasonal under Stata.
Can you help me please ?

Thank you.

[Nick Cox]

I don't see there can be any such thing. For example, one kind of seasonality common with environmental data is a single annual cycle which you can often model well by a regression with a sine and a cosine term. Sometimes more such terms are needed and/or useful. There is no special apparatus there, as it's just regression.

With economic data and regression in mind, it's perhaps more common to have monthly or quarterly data and to use a set of indicator variables in a regression. Otherwise similar comments may apply.

Neither of these is a full-blown time series model. There are many flavours and seasonality is handled differently between them.

So, just in one minute, I've identified three rather different ways of thinking about seasonality.

Otherwise put, tell us about your data: panel or single series; regularly or irregularly spaced; nature of response; nature of predictors; ideas of generating process, and you should have a chance of precise advice.

Please note the request on Statalist to use full real names.

[Mick Kountras]

Hi Nick,

I will modify my profile accordingly asap.
I am working on quarterly data, using regular regression on it.

many thanks,
Mick

[Nick Cox]

I imagine most people would look at indicators for quarters.

[Mick Kountras]

Thank you Nick but could you be more precise, as it is the 1st time I do this .

[Steve Samuels]

I do notice that you missed Nick's and the Statalist FAQ request to use your full real name. This is the way we show professional respect for one another.

[Mick Kountras]

I feel stupid but how can we change our ID ...?

[Mick Kountras]

Someone to help me ... ?

[Alan Riley (StataCorp)]

I feel stupid but how can we change our ID ...?

Click on the 'Contact Us' link at the far bottom right of any forum page. Fill out the forum and let us know your current and desired userid in the message.

[Mick Kountras]

Done!
In the meantime, if someone could help me with this seasonality topic, that would be great ...
Mick

[Nick Cox]

What's the question? What you want to do is just part of an application of regression. I think you are going to need to spell out what's unclear, preferably by telling us about your data and your Stata commands.

[Mick Kountras]

I imagine most people would look at indicators for quarters.

If my understanding is correct, I need to create indicator variables for quarters and regress my series on it.
What is then the criteria to decide wether my series is seasonal or not ?

Many thanks,
Mick.

[Nick Cox]

One criterion (singular) is that one or more of your indicators have effects significant at a chosen level. What else would it be?

[Mick Kountras]

So I create 3 dummy variables, let us say Q1, Q2 and Q3.
I regress my series on it and if for each of these variables, my Fisher statistics is above my confidence level, it means that there is no seasonality ?

Sorry if my questions are obvious but I never treated this topic before.

[Nick Cox]

You seem confused on some basic concepts, test statistic, confidence level and significance level. These go way beyond the specific topic of seasonality, so I recommend closer study of an introductory text.