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

Showing results for

- JMP User Community
- :
- Discussions
- :
- Confidence intervals in prediction profiler (inter...

- 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

A week ago
(71 views)

Hi

Could you help me please? I am still learning DOE.

How can I interpret confidence intervals in the prediction profiler?

So there are significant effects and interactions. However when "discount" is 0 and "direct_comm" is 1 the difference is not significant (picture below)?

So how may I interpret another case when CI of "discount" are significant but CI of "direct_comm" are not?

The remaining:

Big Thanks!

1 REPLY

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

A week ago
(45 views)

There are two kinds of interpretation involved in your question: interpretation of the tests of the parameter estimates and the interpretation of the model predictions.

The three terms in the model are significant. You should not delete any of them or you will bias the model.

The predictions are correct. The traces seen in the Prediction Profiler are based on the model and the given factor levels. The interaction effect (2.5) is almost as large in magnitude as the first-order (or main) effects (3 and -3) so it is a major feature of the profiling.

A significant **interaction effect** means that the *effect of one factor (discount) depends on the level of another factor (direct_comm)*.

- There is practically no effect of changing discount when direct_comm is 1.
- There is practically no effect of changing direct_comm when discount is 0.
- There is a strong effect of changing discount when direct_comm is 0.
- There is a strong effect of changing direct_comm when discount is 1.

This case illustrates why interactions in the response are so important to modeling and prediction. The Prediction Profiler is a great way to understand and exploit interaction effects.

Learn it once, use it forever!