Hello,

I have 2 groups; a patient and a HC (healthy control) group. Performing rsfMRI..

Method 1-
I assessed the ROI's (seed) connectivity against several targets in HC only.
I did this by selecting the HC group in subject effect, rest in condition, the ROI in sources and checked results explorer.
then I selected the patient group in subject effect, rest in condition, the ROI (same as above) in sources and checked results explorer. (with same targets)

I compared the 2 results and the difference is big; In HC, you can see the seed connecting to, lets say, all 20 targets, while in the patient group you can see that the seed is connected to 5 out of the 20.

The HC (n=20)
patient (n=10)

Method 2-
I selected the HC group and the patient group simultaneously in subject effect, rest in condition, the ROI in sources and checked results explorer. 

This time the results only showed 1 ROI to be significant.

May i know what contributes to lack of significant difference when i combine and when i don't? even thought in HC, for example, the seed was "highly connected", as was shown in the colour bar, to one of the targets, and was not connected at all in the DAI (when done separately), but when i combined this seed/target didn't show that it was significantly different between patient and HC!

Any help?

Hi Walid,

Method 2 is the proper way to test for between-group differences (I am assuming that when you select both the HC and patient groups simultaneously you are then entering a between-subjects contrast [-1 1] to look at the between-group differences).

There are many ways in which differences in the significance of individual groups do not translate to significant differences in the between-group comparison. Even if the average connectivity in the two groups between two ROIs is exactly the same (so basically no between-group difference in connectivity at all), you could still see differences in the significance of the individual within-group results, for example, if the Patients group has higher inter-subject variability (that would lead to more "significant" results in the HC group than in the Patients group), or if the Patients group is smaller (that would also lead to more "significant" results in the HC group). In general one should always be careful when interpreting differences in significance between two groups (because those differences may arise from differences in power between the two analyses, rather than true differences in connectivity between the two groups) and should use instead the proper between-group comparison test when wishing to make inferences about between-group differences. 

Hope this helps
Alfonso

Originally posted by Walid Yassin: