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
- :
- Re: Analyzing results when interaction is not possible

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
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 10, 2019 4:20 AM
(148 views)

Hi!

A newbie question.

I have a basic experiment with 2 factors:

The problem is that a combination of X1 and X2 technically is impossible.

Can I still analyze the results of first 3 patterns using regression? Or is there a differnt approach to this problem?

Actually this is so called ABN test on a site, Original (A), new feature version B(X1), version С(X2) the same new feature as B(X1) but is displayed to a user in a differnt way.

Learning DOE

1 ACCEPTED SOLUTION

Accepted Solutions

Highlighted

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

The data determine what is estimable. Omitting one of the runs will limit the model that can be fit. It is OK, though. Remove the term that would cause the singularity. For example, if you omit either the (-1,-1) or the (+1,+1) run, then you must remove the interaction term from the model or else it will cause a singularity in the linear regression. So the nature of each run is important in defining which effect is estimable.

Remember that with only three runs, it is possible to estimate the intercept and the two main effects. It is impossible to perform any hypothesis tests on the model as a whole or on individual terms. So the number of runs is important in defining which tests are possible.

Learn it once, use it forever!

1 REPLY 1

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

The data determine what is estimable. Omitting one of the runs will limit the model that can be fit. It is OK, though. Remove the term that would cause the singularity. For example, if you omit either the (-1,-1) or the (+1,+1) run, then you must remove the interaction term from the model or else it will cause a singularity in the linear regression. So the nature of each run is important in defining which effect is estimable.

Remember that with only three runs, it is possible to estimate the intercept and the two main effects. It is impossible to perform any hypothesis tests on the model as a whole or on individual terms. So the number of runs is important in defining which tests are possible.

Learn it once, use it forever!