Re: [Freesurfer] Interaction test

yes, where "stronger" means greater in magnitude ignoring the sign doug

On 2/17/14 10:41 AM, Yang, Daniel wrote:

I see. You are saying that the interpretation can be made in two ways:

blue/cyan =

the cortical thinning effect of Age is stronger in patients (vs. controls)

Or, the cortical thickening effect of Age is stronger in controls (vs. patients)

Thanks, Doug! Daniel

-- Daniel (Yung-Jui) Yang, Ph.D. Postdoctoral Researcher Yale Child Study Center New Haven, CT Tel: (203) 737-5454 E-mail: _yung-jui.yang@yale.edu_

On 2/17/14 9:55 AM, "Douglas Greve" <greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu> wrote:

Yes, you should see cyan. Note that you would see cyan if group1
slope is 0 and group 2 slope is positive. Probably not a problem
with age, but it is just something to think about.
doug




On 2/17/14 1:38 AM, Yang, Daniel wrote:

...

Dear FreeSurfer developers and experts,

If I have a binary factor (Group1: patients, Group2: controls)
and a continuous variable (age), I know I can follow Contrast2 in
the following wiki to test the interaction effect.

http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf2G1V

Contrast2: [0  0  1  -1]  tests if Age slope in patients (Group1)
> controls (Group2), and if Age slope_Group1 - Age slope_Group2 >
0, I can see red/yellow.
For example, cortical thickening is accelerated in patients (vs.
controls).

I am wondering how I can test for whether cortical thinning is
accelerated in patients (vs. controls)?

Here, the slope for Age would be negative. And if cortical
thinning is accelerated in Group1 (vs. Group2), I should observe
Age slope_Group1 < Age slope_Group2, that is, Age slope_Group1 -
Age slope_Group2 < 0, thus, I should expect to see blue/cyan?

Best,
Daniel

-- 
Daniel (Yung-Jui) Yang, Ph.D.
Postdoctoral Researcher
Yale Child Study Center
New Haven, CT
Tel: (203) 737-5454
E-mail: _yung-jui.yang@yale.edu_


_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer