Dear Alfonso and all,

I have some question about the ROI-to-ROI FC analysis 2nd-level results explorer. I have three group: HC, MCI and AD. 17 spherical ROIs of a certain network.
1) If I don't have any hypothesis and want to look into the FC group comparison of all possible connections between HC and AD, controlling for age and gender, I select the subject effects of [HC, AD, age, gender] and set the contrast [1 -1 0 0 0]. Then should I use the multiple correction in the result explorer of "FDR seed-level" or "FDR analysis-level" ? (the default option is "seed-level" ) How many connections these two methods will corrected for exactly? For the 17 ROIs, is it 17 for seed-level and 136 for analysis-level?

2) If I want to look at the FC connections only within one group, e.g. HC group, should I also control for the covariates like age and gender, etc.? (select "HC" and set contrast [1] or select "HC, age, gender" and set contrast [1 0 0]?) I tried both ways but the former showed many significant connections (use "FDR analysis-level" correction) and the latter showed no significant result.  

3) I found the result are different when analysing the FC within the HC group by using these two ways: a) subjects includes HC and AD group and select only "HC, age, gender" and use contrast [1 0 0]; b) subjects includes HC, MCI and AD group and select the same subject effect and use the same contrast. The difference is that the 1st-level covariate of HC including more 0 for MCI subject in the second way and the age, gender covariates also include more values of MCI subjects. I have checked that the FC correlation coefficient for the two methods are exact the same, but the result in the 2nd-level explorer differs a lot. The former way showed more significant result. Do you have any idea why it was different and what is the better way?

4) I usually choose to mask the ROIs with the grey matter (GM) mask in the "setup" step, but now I found for one of the analysis there is rare significant FC connection when masking with GM mask than not masking with GM mask. It maybe because the GM mask is not so precise and it cut some voxels off the ROI and get rare result, or the ROI may includes some white matter voxels. What do you think about this? Is it OK to do the analysis without masking with GM mask?

I greatly appreciate for your help!
Best regards,
Yifei