Hi Ariel,

Sounds like you want to include more PPI regressors to make this a generalized PPI model (e.g., McLaren et al., 2012, NeuroImage). That would mean your first-level model would include at least four additional EVs: dlPFC*Load_1, dlPFC*Load_3, dlPFC*Load_5, and dlPFC*control. The gPPI model would also let you test whether connectivity varies as a function of load.

The group-level model sound reasonable from the description, but I can't be sure without seeing the actual model.

Cheers,
David

> On May 3, 2016, at 5:55 PM, Ariel Eckfeld <[log in to unmask]> wrote:
>
> Hi all,
> I was hoping to have some help in making sure I've set up my PPI models correctly as the results we have been getting seem somewhat suspicious. This is a parametrically varied spatial working memory/capacity task that consists of 4 different working memory load conditions (1, 3, 5, & 7). I am running a PPI to examine the difference between my patients and controls at the highest working memory load (7) using working-memory relevant ROIs (e.g., dlPFC). Briefly, the first level models have text files for load 7's task onset times, the timecourse for the ROI from the subject, the interaction between the two, and control EVs for the other loads' onset times (1, 3, 5). Group models are split into main effects and an age x group interaction, and both covary for age, gender, handedness, and percent correct responses on the task. Unfortunately screenshots are too large to attach here, so let me know if more detail would be helpful.  I appreciate any comments or suggestions!
> Thanks!
> Ariel