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
- :
- Urgent question: interaction effects without main effect

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

Aug 31, 2009 9:07 AM
(1306 views)

I encountered one problem when I try to include an interaction factor in one model.

JMP doesn't allow me to include the interaction effect without the main effect. It looks like the model must include both main effects if I want to have the interaction effect in the model, for example, JMP accept y=a+b+a*b, but not accept y=a+a*b

anybody can answer my question?thanks.

JMP doesn't allow me to include the interaction effect without the main effect. It looks like the model must include both main effects if I want to have the interaction effect in the model, for example, JMP accept y=a+b+a*b, but not accept y=a+a*b

anybody can answer my question?thanks.

3 REPLIES

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

JMP insists you maintain hierarchy within the model, therefore you cannot have an interaction effect without the main effect being present. Once the model is produced with both effects present you may find that the main effect is small in comparison and could be excluded but its presence is required whilst building the model.

Therefore build your model maintaining hierarchy and check significance of each effect. If the model itself is insignificant, you could consider stepwise regression to eliminate highly insignificant terms.

Therefore build your model maintaining hierarchy and check significance of each effect. If the model itself is insignificant, you could consider stepwise regression to eliminate highly insignificant terms.

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

Hi Cowboy,

Which version of JMP are you using? I tested it quickly in JMP 7 and JMP 8 and if you leave out a main effect that's used in a interaction JMP will notify you that the effect is missing but it does go ahead and fit the model.

Can you tell us exactly what's happening and what message you're receiving?

Jeff

Which version of JMP are you using? I tested it quickly in JMP 7 and JMP 8 and if you leave out a main effect that's used in a interaction JMP will notify you that the effect is missing but it does go ahead and fit the model.

Can you tell us exactly what's happening and what message you're receiving?

Jeff

-Jeff

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

Not a JMP comment, but a statistics one.

By including an interaction in the model without the main effect, you are saying that the missing main effect is exactly 0. To the extent that it is greater than 0, you are contaminating (confounding) the interaction with the missing main effect.

Caveat emptor!

By including an interaction in the model without the main effect, you are saying that the missing main effect is exactly 0. To the extent that it is greater than 0, you are contaminating (confounding) the interaction with the missing main effect.

Caveat emptor!