Hi
Your confusion as to how to interpret your 'activation' seems related to a common misconception: areas which survive thresholding for C3 are areas where G1-G2>0, i.e. where the difference is significantly larger than zero. This does _not_ require C1>0 and/or C2>0! E.g. it could be that in C1 these areas are mildly positive (but sub-threshold) whereas in C2 they are mildly negative - the overall difference then could very well be large enough for C3>0.
Another posibility is that both C1 and C2 are negative, but C2 significantly more so. The way to check is to (i) add the negative contrasts (e.g. -1 0) to check for de-actiations and/or to investigate explicitly within your clusters using FEATquery.
hth
Christian
On 19 Jan 2011, at 04:51, Susan wrote:
> Hi everyone,
>
> I've been working on a PPI analysis and I would love some input regarding a between-group analysis result.
>
> I used a functionally defined ROI for the seed, and ran a PPI on 40 subjects (lower-level). Then I input these 40 .feat folders into a higher-level analysis with 2 groups (20 subjects per group). I defined EV1 as Group1, and EV2 as Group 2, and setup the following contrasts:
>
> C1 Grp 1 mean: 1 0
> C2 Grp 2 mean: 0 1
> C3 Grp 1>2: 1 -1
> C4 Grp 2>1: -1 1
>
> Here's where I'm stuck. The results show that nothing survives thresholding for either contrast 1 or contrast 2. If I had stopped here, I would interpret this to mean that there are no regions that show task-specific correlation to the seed ROI at the current threshold. However, I'm confused because contrast 3 shows clusters that survive the same threshold.
>
> I want to interpret C3 to mean that Grp 1 shows greater task-specific correlation in these areas than Grp2. But how do I interpret significant between group differences when there were no regions that correlate with the seed ROI in either group alone (C1 and C2)?
>
> Thanks for any thoughts/input.
>
> Susan
|