turn on suggestions

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

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Forecasting
- /
- Wide Confidence Interval for pedictions SAS foreca...

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
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-04-2016 02:51 PM

what can be the reason for having very wide confidence intervals for the predictions? Should one report these predictions, if yes how would he explain the wide C.I?

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

01-04-2016 03:07 PM

How wide is "very wide"?

Wider intervals generally come from the variability in the observed data for the number of records used. Generally a small sample will have a larger CI for a given dispersion of data than a larger sample with the same dispersion.

And with forecasts, the farther from the observed data in time the wider.

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

01-04-2016 03:21 PM

Wider Prediction Intervals (PI) may indicate that your forecast model isn't optimal.

Should you report predictions? That depends on if they're useful in a practical sense, given the PI. If they're too variable to be of use, then they're not, and the criteria for such a definition would vary from field to field. In the medical field, very small PI would be optimal, while in a retail setting wider PI may be perfectly fine.

Should you report predictions? That depends on if they're useful in a practical sense, given the PI. If they're too variable to be of use, then they're not, and the criteria for such a definition would vary from field to field. In the medical field, very small PI would be optimal, while in a retail setting wider PI may be perfectly fine.

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

01-04-2016 03:35 PM

Hello -

Confidence limits are based on the prediction standard errors and a chosen confidence limit size. The variance of each forecast is called the prediction error variance, and the square root of the variance is called the prediction standard error. The variance is computed from the forecast model parameter estimates and the model residual variance. Wide CI usually indicate either a poor performing model or lots of noise in your data.

Have you considered using forecast combinations, also called ensemble forecasts? They are a known technique for improving forecast accuracy and reducing the variability of the resulting forecasts. See: http://blogs.sas.com/content/subconsciousmusings/2014/06/18/combined-forecasts-what-to-do-when-one-m...

Thanks,

Udo