Print

Print
Hi Ekarin,

It doesn't matter if the conditions are related or not. The key is that in either model, the PPI terms fully model all possible connectivity values of the task. With 4 conditions, this is most easily done with 4 task EVs, 4PPI EVs, and the physiological EV.

In the gPPI model, each PPI EV is relative to the baseline connectivity (whatever task is not modeled), not the other tasks as the other tasks are included in the model. In fact, if you look at the paper, only the gPPI model captures the true connectivity values with more than two conditions.

Best,
Donald

Best Regards, 
Donald McLaren, PhD


On Mon, Mar 19, 2018 at 4:15 PM, Ekarin Pongpipat <[log in to unmask]> wrote:
Thank you, McLaren, I had considered using gPPI, but was worried that our conditions would be too related and would "lose" some PPI effect by modeling the other conditions as zero. Specifically, our conditions are 0-back, 2-back, 3-back, and 4-back of the n-back paradigm and if I were to model these conditions using gPPI then, for example, the PPI for the 2-back would be saying that is the change in the relationship between the seed and voxel between the 2-back and the other conditions (which might not be ideal since the 3-back and 4-back are related to the 2-back)? Whereas if I were to model, for example, 2-back minus 0-back, the PPI term would now be the change in the relationship between the seed and voxel between the 2-back and the 0-back? Or if valid, it would be ideal to have one EV be a linear effect. Would this logic and approach be valid/appropriate given our n-back conditions?


On Mon, Mar 19, 2018 at 11:25 AM MCLAREN, Donald <[log in to unmask]> wrote:
Hi Ekarin,

If you read my 2012 paper on generalized PPI, you can absolutely model 4 Condition EVs, 4 PPI EVs, and the physiological EV.  If you model each condition separately, then you will have much more freedom on the contrasts that you can generate. There is also a nice paper, with video, for doing this in FSL (Harrison et al., 2017; https://www.ncbi.nlm.nih.gov/pubmed/29286444)

Best,
Donald

Best Regards, 
Donald McLaren, PhD


On Mon, Mar 19, 2018 at 12:18 PM, Ekarin Pongpipat <[log in to unmask]> wrote:
Thank you, Anderson! I was able to figure out the convolution part.

For the linear contrast question:
Rather than making a model with 4 EVs of cond1-control, cond2-control, cond3-control, and cond1+cond2+cond3+control

Would it be appropriate to make the following EV?
Linear: -1.5*control - 0.5*cond1 + 0.5*cond2 + 1.5*cond3

If so, would it need to include the following EVs as well to make it complete?
Average: .25*control + .25*cond1 + .25*cond2 + .25*cond3
Quadratic: -.5*control + .5*cond1 + .5*cond2 - .5*cond3

I hope this is more clear.

Thank you again!,
Ekarin

On Sun, Mar 18, 2018 at 4:21 PM Anderson M. Winkler <[log in to unmask]> wrote:
Hi Ekarin,

Please see below:


On 12 March 2018 at 13:58, Ekarin Pongpipat <[log in to unmask]> wrote:
Hi FSL experts,

I had several questions about creating the PSYCH term (i.e., a contrast; https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PPIFAQ). If I have four related stimuli conditions in a block-design, is it possible to make the following contrasts within the same model: cond1-control, cond2-control, cond3-control?

Yes, create these 3 EVs with the differences, and include a 4th EV that is cond1+cond2+cond3+control.
 
Would it be appropriate to convolve this contrast?


Yes, it is appropriate to do the convolutions on these EVs (not contrasts) before or after the subtractions, and also multiply by the physiological measurement.

 
If so, how would you determine the duration of the convolution (note: block duration for cond1, cond2, and cond3 are the same but the block duration for control is much shorter).

I don't understand this question.
 

In a different PPI model, is it possible to model the PSYCH term as a linear contrast? If so, would I need to also include an average and quadratic contrast to make the model complete? And similarly, would the convolution be appropriate here as well?

I don't understand this question either, sorry.

All the best,

Anderson
 

Thank you!,
Ekarin

--
Ekarin E. Pongpipat, M.A.
Ph.D. Student in Cognition and Neuroscience
Cognitive Neuroscience of Aging Laboratory
Center for Vital Longevity
School of Behavioral and Brain Sciences
The University of Texas at Dallas

--
Ekarin E. Pongpipat, M.A.
Ph.D. Student in Cognition and Neuroscience
Cognitive Neuroscience of Aging Laboratory
Center for Vital Longevity
School of Behavioral and Brain Sciences
The University of Texas at Dallas