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
- :
- Time Series Analysis Predicting Future Values

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
- Permalink
- Email to a Friend
- Report Inappropriate Content

Feb 29, 2016 3:57 AM
(3223 views)

Dear All,

Not doing any time series analysis before I've been asked to predict the 2016 Sales as week by week and as final amount.

The company is a seasonal leisure and tourism (more exactly a resort hotel).

The sales end by the 44th week of the year (hotel close date) and starts 73 weeks before the hotel close date.

To fit the sales data more to a non seasonal annual sales data I only took the last 52 weeks of sales.

I have a data set of last 4 years sales.

But I am not aware of how to shape the data like; should choose Time Series of Column as annual or weekly?

Should choose Seasonal ARIMA (0, 1, 1)(0, 1, 1)12 or Seasonal ARIMA (0, 1, 1)(0, 1, 1)52?

Or since the data is to be considered as annual should I choose ARIMA?

How to fill up the predicted weekly values into data table?

Tables are attached and would appreciate any help.

Solved! Go to Solution.

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Feb 29, 2016 9:58 AM
(4661 views)

Should choose Time Series of Column as annual or weekly?

If you want to predict weekly sales you should use weekly data. Besides, there are only four years of data, so it’s not enough to run a time series analysis on yearly data.

Should choose Seasonal ARIMA (0, 1, 1)(0, 1, 1)12 or Seasonal ARIMA (0, 1, 1)(0, 1, 1)52?

If a seasonal pattern occurs in a 52 week cycle, which appears to be shown in your data, yes use 52.

Or since the data is to be considered as annual should I choose ARIMA?

See my comments above.

How to fill up the predicted weekly values into data table?

First, on the Time Series dialog choose how many # of periods you want to forecast. The default value is 25. In your case, enter 35 (to fill up through 2016) or more. Second, after you’ve built seasonal ARIMA models, click on the red triangle and select either *Save Columns* or *Save Prediction Formula* to obtain a new JMP table that contains forecast.

http://www.jmp.com/support/help/Modeling_Reports.shtml#104953

I also suggest that you use a Log transformation on your sales to ensure positive predictions, and then use Exp (or 10^) to backtransform.

1 REPLY

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Feb 29, 2016 9:58 AM
(4662 views)

Should choose Time Series of Column as annual or weekly?

If you want to predict weekly sales you should use weekly data. Besides, there are only four years of data, so it’s not enough to run a time series analysis on yearly data.

Should choose Seasonal ARIMA (0, 1, 1)(0, 1, 1)12 or Seasonal ARIMA (0, 1, 1)(0, 1, 1)52?

If a seasonal pattern occurs in a 52 week cycle, which appears to be shown in your data, yes use 52.

Or since the data is to be considered as annual should I choose ARIMA?

See my comments above.

How to fill up the predicted weekly values into data table?

First, on the Time Series dialog choose how many # of periods you want to forecast. The default value is 25. In your case, enter 35 (to fill up through 2016) or more. Second, after you’ve built seasonal ARIMA models, click on the red triangle and select either *Save Columns* or *Save Prediction Formula* to obtain a new JMP table that contains forecast.

http://www.jmp.com/support/help/Modeling_Reports.shtml#104953

I also suggest that you use a Log transformation on your sales to ensure positive predictions, and then use Exp (or 10^) to backtransform.