Hi conn community,
I will be very appreciate your help regarding to secon level voxel-to-voxel analysis. 

I performed this analysis on 10 subjects (6 conditions) and have got strong negative contrast. What does really the negative contrast means for the PCA? What the blue bolbes tell us about connectivity differences across subjects (as was stated in conn manual)? for the PCA? 

Moreover, as far as I understand, dimensionality reduction lead to loosing some information about variability of connectivity patterns. How we can use connectome_MVPA_2 (3...n)? What this information tell us from practical point of view?

Thank you

Hi Alla,

First one observation, because PCA analyses are rotation-invariant the particular sign of the resulting contrasts in MVPA analyses carries no information (because of this in general it is recommended that one uses either F-tests or two-sided T-tests on your second-level analyses when analyzing MVPA results, instead of one-sided T-tests). 

Significant 'blobs' in these results mean that there are significant differences in the functional connectivity between these blobs and the rest of the brain (differences across conditions or across subjects, depending on how you set up your second-level analyses). For example, if you select two conditions in your results (and enter a [1,-1] contrast), and then select several of your MVPA_1 to MVPA_n measures in the 'measures' list (and leave the default eye(n) contrast there), after clickin on the 'explore voxel-to-voxel results' button the resulting 'blobs' will indicate which areas show different in connectivity patterns (connectivity between these areas and the rest of the brain) when comparing the two conditions. In order to clarify what these 'differences between condtions' mean, you could then perform post hoc analyses exploring the connectivity between each of these blobs and the rest of the brain. For example, let's say you find a significant blob in the amygdala in your original MVPA analyses (indicating significant differences in amygdala functional connectivity between conditions). You could then use this amygdala blob as seed in post hoc seed-to-voxel analyses, and explore in more detail what particular aspects of the functional conneciivty with the amygdala might differ across conditions. Let me know if this clarifies.

And regarding dimensionality reduction, the number of chosen MVPA components affects your sensitivity to find significant across-conditions or across-subject effects. Given your relatively low number of subjects I would probably recommend using only two MPVA components (simply choose the MVPA_1 and MVPA_2 components when performing the second-level analyses, it does not matter whether you performed the first-level analyses using a larger number of components). In general, using too many components could degrade the sensitivity of your second-level analyses (simply because of low power for any given sample size), and using too few components could also degrade the sensitivity (because you could miss smaller variations in connectivity across subjects or across conditions which do not get to be represented by the limited number of components chosen). As a rule of thumb in general I would probably recommend using either 1:5 or 1:10 ratio of number of MPVA components to number of subjects.

Hope this helps
Alfonso




Originally posted by alla yankovskaya:

Hi Alfonso,
Very much appreciate your clear and understandable explanation, even though that my question did not directly related to technical issues of conn.
Working with conn for the last 6 month. Great toolbox! Thank you

Dear Alfonso,
Regarding second-level voxel-to-voxel analysis. I have 6 conditions. I am interested in following connectivity patterns: condition 1 by itself, condition2 by itself, condition 3 by itself and contrast [1,-1] between conditions 1 and 2, and 1 and 3. In subject effect list I selected 'all', in condition I left 1, in measures list i selected connectome 1&2. 

I did not get any significant results for conditions 1 and 2 by itselves.In contarst, condition 3 showed some areas connected significantly to the rest of the brain. Please, could you help me understand what the absense of any significant (I mean, corrected_FDR) patterns for conditions by itselves does mean? Or I enterred wrong contrast (1)? 
Thank you
alla