Hi Athena,
When you say 'this time differs when I compare it" I presume you mean that it differs upon visual inspection. At this stage you have not tested for statistical significance of this 'difference'. One reason why your comparison might not reach significance is that the within-group variability is too large.
hth
Christian
On 21 Apr 2011, at 16:33, Athena Demertzi wrote:
> Dear FSL people,
>
> another issue I am baffled over is the following:
> I would like to compare resting data of controls and patients. For this reason I do the following: I order a GICA temp concatenation with both patients and controls, then I build the design matrix to model my two groups and I do dual regression with permutation testing. The resulting p-corr maps for the control>patients, patients>controls, mean controls, mean patients are derived. I use a template to select my component of interest (e.g. DMN) and then I check the resulting maps: I realize that my controls and patients do not differ and this was against our hypothesis. Then, I went to see what is the case in the patients group only: I order GICA temp concatenation and I do dual regression on this output. The mean effect of patients this time differs when I compare it to the mean effect from the GICA with controls and patients together (more specifically, in the GICA with patients and controls I found a preserved DMN; but in the GICA with patients only no DMN is detected).
>
> What does this mean? And how should I best proceed with the comparison of my two groups?
>
> Thanks in advance for your help!
>
> Best
>
> Athena
|