# Interpreting effects of gPPI analysis

Hello, AFNI mindhive –

I’m currently analyzing a set of fMRI data and am running into some trouble implementing a gPPI analysis. My apologies in advance if a similar question has already been posted – I couldn’t find anything when I ran a search.

(Some background if it’s useful: We’ve got a 2x3 design, and we functionally defined seed regions based on sensitivity to the 2-level factor. We’re interested in examining functional connectivity differences as a function of the 3-level factor.)

I’ve worked through Gang Chen’s very helpful guide on implementing gPPI and have run my deconvolution analyses for each subject. Here’s where I’m having trouble:

[ul]
[li] I’m not sure whether to be running the ANOVA on the beta values or the r values. Step 6 of the guide says to output the r coefficient in the deconvolution because the r value will be used in the group analysis, but Step 7 says to run the group analysis on the beta coefficients.
[/li][li] Once I’ve run the ANOVA and identified regions whose functional connectivity to the seed varies based on the task, I’d like to get a sense of which conditions involve greater connectivity and which ones involve less. To do that, I assume I need to extract values from the regions using something like 3dmaskave and examine the pattern of activity. I’m unsure whether it would be appropriate to extract beta values or r values (and, if the latter, what sorts of conclusions I can draw from the values I extract).
[/li][/ul]

Sahil

Hi, all,

I just wanted to follow up on this question – I figure it might have gotten buried since I posted it on a Friday afternoon!

Thanks!
Sahil

You can theoretically take either betas or correlation coefficients to the group level (though in the case of the latter, you would probably want to z-score). As I understand it, it is more common to do group statistics over individual subject betas. See the bold text in Gang’s post here: https://afni.nimh.nih.gov/sscc/gangc/PPI.html

Regarding the issue of understanding directionality, I’ve used the approach you describe (3dmaskave over a region, extracting betas in my case) and have interpreted positive vs. negative correlations on the basis of the sign of the fit coefficient. However, if you would prefer to do this with z-scored correlation coefficients, I don’t personally see any reason you can’t. Alternatively, you could consider running a direct comparison over two different PPI beta maps (e.g., run 3dttest++ on the beta maps for factor A vs. factor B).

Hope this helps! I’ve only recently run PPI analyses, so someone else please reply if this isn’t accurate.