Hello,

I ran a gPPI analysis and have written up my results and made figures, but I want to make sure I'm providing accurate information and have a few questions:

• What are the appropriate effect sizes/t values to report and how do I get those values? I've read other posts on this forum like this one, but the screen shots are from an older version of CONN and with the gPPI aspect of the analysis I'm not sure what belongs in my results table. I can see that the p values are different from those of the original results and it's my understanding that the graphed REX analysis is running a different statistical analysis than the original. When reporting results, do I use the t values from the REX results and the p values from the original clusters? How do I find the effect sizes that are plotted in the REX bar graph, and if I report them should I report the difference between the conditions if that's what I'm interested in? I'm attaching screenshots of my REX and original results, if that helps at all.
• For some of my second level analysis I get this "WARNING: possibly incorrect model: non-estimable contrasts (suggestions: simplify second-level model)." My second level model includes the AllSubjects constant, SexFemale, SexMale, Age, MedsOn, and MedsOff - so I have the continuous variable age (which I orthogonalized) and then two binary variables each for sex and whether or not the participant was taking medication. I am looking at the "Effect of AllSubjects" because I want to control for sex, age, and medication use. For this analysis I'm contrasting the difference of the experimental > control conditions [-1 1]. Is my analysis set up incorrectly and I should only have (for example) a "Sex" binary variable and not "SexFemale" and "SexMale"? Is this warning just something I can ignore since I do have significant results, or does this warning mean my results are inaccurate?
• My experiment consists of 3 runs with 9 sections each that are either the rest, control, or experimental conditions. I am not interested in the rest condition, so I included effect of rest, control, and experimental conditions in my denoising step then only put the experimental and control conditions (which I want to contrast) in my gPPI bivariate regression analysis. Is this the correct set up for this design since I have 9 periods of rest (where subjects just fixate on the screen) interspersed with the control and experimental conditions in my task, but I'm not actually interested in analyzing rest in this analysis?

(I also want to say thank you for the advice I got on my previous post about QA - that made me feel a lot better about my data so thank you).
-Rachel