Hi George, We generally use option one for most analysis focusing on comparisons of the groups. But it depends on what the aim of your project is. Shruti. On Apr 16, 2015 11:08 AM, "George Chahine" <[log in to unmask]> wrote: > Dear all, > > I have a sort of basic question about group ICA design. I wish to run ICA > on two groups (control vs patients). We have the following options > concerning the group analysis: > > 1) do the initial ICA definition on all subjects regardless of group, and > then dual regression on each subject in the analysis. > 2) do the ICA definition only on the control group, and then dual > regression on the subjects from both groups. This will decrease false > negatives. > 3) do the ICA definition on a large control group (that does not include > the control subjects in the subsequent analysis) and then dual regression > on all subjects that will be part of the analysis. > 4) do ICA on the patients alone and the control group alone, and then do > dual regression for each group separetely and carry the second level > analysis on matching networks. > > 4 is probably the least preferred option, as it will introduce too much > bias and false positives. 3 to me seems the best, but I am not sure if it > is valid to do the initial ICA on subjects outside the main analysis. > > I would greatly appreciate your help on this issue. > > Best regards, > > George >