Print

Print
The first one is almost correct.

EV2 should have zero mean. Then mean of your scores is 33. So, EV2 should be 1 -3 -1 3.
In this way EV2 will capture any differences over and above the mean activity that are attributable to you scores.
without EV1, your maps will contain mean effects as well, and you will not be able to dissociate what is attributable to the score variation across subjects.
At least, this is my understanding.

Hope this helps.

BRad


On Oct 25, 2006, at 8:09 AM, Dharol Tankersley wrote:

FSL helpers:
 
I have two questions.
First, when setting up a higher level analysis to see how activation for a given contrast covaries with an additional EV, for example, scores on a pyschometric test, Is it necessary to have two EVs in the model -- one that is all ones, and the other wth the scores of the psychometric test?
For example,
 
            Group    EV1    EV2
Input       1            1        34
Input       1            1        30
Input       1            1        32
Input       1            1        36
 
What would happen if you only put in the EV of the scores?
For example,
 
            Group       EV1
Input       1              34
Input       1              30
Input       1              32
Input       1              36
 
 
Second, if an EV correlates negatively with ROI activation in a given contrast, does the model need to be altered in any way to make sure it identifies the correlation?
 
Thanks,
 
Dharol
 

Dharol Tankersley
Dr. Huettel's Labratory
Brain Imaging and Analysis Center
Duke University
919-681-9935