help > RE: PPI Interpretation
Feb 24, 2019  12:02 PM | Donald McLaren
RE: PPI Interpretation
Originally posted by Phil Kuhnke:
Dear all,

I have a couple of questions regarding the interpretation of PPI results.

- A significant PPI effect in voxel Y with seed region X for task A > B means that the regression slope between the time courses of X and Y is significantly higher during A than B, above and beyond the activation of both X and Y for A > B, and their task-independent correlation, correct?
>>>> Almost, "above and beyond the activation of both X and Y for A > B, and their task-independent correlation" should be "above and beyond the correlation of the X and Y, the activity of A and the activity of B in voxel Y".
Now, does this mean that we would find a significant PPI effect if the correlation between X and Y during B is highly negative, and it becomes less negative or even 0 during A? Can we really interpret such effects as an increase in coupling / functional connectivity between X and Y for A > B? It seems to me that the coupling even rather decreases (from an anti-correlation to no correlation).
>>>> PPI specifically relates the slopes. One could limit the analysis to only voxels that the correlation with A or B is positive.

- Related to this, aren't strong negative correlations also interesting, in that they could reflect inhibitory influences between brain regions? Maybe what we should truly analyze is the absolute regression slope?!
>>>> Yes. One could take the absolute values. I'm not sure if you can use absolute values in a contrast though.

- My PPI analysis revealed that seed region X (6mm sphere around group activation peak for task A > B) couples more strongly with region Y during A than B, although Y was not (significantly) activated for A > B in the group univariate analysis. How can such a result be interpreted? How is it possible that X and Y functionally interact more strongly during task A than B, although Y shows no significant activation differences for A > B?
>>>> There could be a number of explanations. For example, the variance across subjects is very high for the task and thus the effect is not significant, but the coupling change is highly consistent and thus is significant. It could also be due to how you modeled the task or connectivity as well. I tend to think of the each brain region having two components to it's activity: (1) the local stimulus; and (2) the connected inputs from other areas. 
>>>> Hope this helps.

Thanks a lot in advance and all the best,

Threaded View

Phil Kuhnke Feb 24, 2019
Phil Kuhnke Feb 24, 2019
RE: PPI Interpretation
Donald McLaren Feb 24, 2019