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
- :
- Help with Regression

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

May 3, 2017 6:32 PM
(1467 views)

I am comparing insulin resistance (continuous variable) between 3 groups of patients, but body mass index (continuous variable) affects insulin resistance. Using JMP, how do I compare these three groups by 'adjusting' for the effect of BMI on the groups (and thereby removing its effect on the comparison) because the BMI is significantly different between the three groups (by ANOVA).

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

May 3, 2017 7:18 PM
(2757 views)

That final statement would have a p-value of 0.0157 (the line on the Main Group under Effect Tests). The effect of BMI has already been removed. The blank spaces should be the Least Squares Means, which you don't have on your report right now. You can get them from the red popup triangle by Response HOMA-IR and choose Estimates > Multiple Comparisons. That will also give you group by group comparisons to declare (as an example) that Group A is different from Group B, etc.

A big caution though. This model has a significant lack of fit. You should investigate and remove that as it is an indication that something is likely missing from the model (maybe the interaction?). Please remember that there is much more to an analysis than just that one line summary that you asked for.

Dan Obermiller

4 REPLIES

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

May 3, 2017 6:47 PM
(1462 views)

You will need to use Fit Model. Y is your insulin resisance. You have two X terms to put into the model: the groups and BMI. You would also likely want to put in the interaction between those two terms. You will see an Effect Tests report that will tell you the significance of the groups. You will also get tests for the significance of BMI and the interaction term.

Dan Obermiller

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

May 3, 2017 6:53 PM
(1458 views)

Thanks so much. It shows that the BMI is highly significant (figure attached). How do I now compare HOMA-IR between the three groups (and remove the effect of the BMI)

My final statement would say.

When adjusted for BMI, the insulin resistance between the subgroups was ____ vs ____ vs _____ (p=0.xx)

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

May 3, 2017 7:18 PM
(2758 views)

That final statement would have a p-value of 0.0157 (the line on the Main Group under Effect Tests). The effect of BMI has already been removed. The blank spaces should be the Least Squares Means, which you don't have on your report right now. You can get them from the red popup triangle by Response HOMA-IR and choose Estimates > Multiple Comparisons. That will also give you group by group comparisons to declare (as an example) that Group A is different from Group B, etc.

A big caution though. This model has a significant lack of fit. You should investigate and remove that as it is an indication that something is likely missing from the model (maybe the interaction?). Please remember that there is much more to an analysis than just that one line summary that you asked for.

Dan Obermiller

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

May 3, 2017 7:23 PM
(1445 views)

Thanks so much. I totally understand. This is just a representative sample out of a bigger dataset with more variables. I just wanted to know where to start looking.