turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Finding seasonality in a data set that has inflation

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 19, 2013 3:38 PM
(1785 views)

I'm looking at a dataset like this:

http://www.census.gov/retail/marts/www/download/text/adv72200.txt

I want to determine if there is any seasonality to the set. The AutoCorrelation shows no spikes but I think it's a little goofy because the data has such a strong upward trend. I'd like to control that upward trend (ie inflation) and then look for seasonality.

Can anyone help me with this?

Thanks in advance!

-Matt

2 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Matt,

You can control for the upward trend by differencing. This is easily handled in the Time Series Analysis module by selecting the difference option. Or, if you prefer, you can calculate the difference yourself by creaing a new column and entering the formula food - food*Row()-1 (i.e., *Row()-1 is subscripted). This subtracts the previous observation from the current observation, which removes trends. The first row wild return an error; just click OK.

The differenced values should no longer have an upward trend.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Apart from the time series approaches-

Fit model can be used with an overall trend data as a fcn of time, (maybe time^2)

With the strong overall trend defined, you can find a function for the overall (linear, polynomial etc) using fit model, then "save residuals"- and examine the residuals by month independent of the trend.

I have seen approaches (sometimes used when time series are not regular steps), of calculating the "phase" of the year in radians (month / 6 * pi), calculating a SIN and COS of this phase, and adding them as terms in the fit model.

regards-