Dear Amy,
not sure whether you extracted beta/contrast estimates or percentage signal change. In any case, it's difficult to tell whether something is a "good variation" or not except you have a clear hypothesis about the values. There might be very small but highly consistent and/or highly significant differences between conditions.
Negative values mean that there's a negative correlation with the expected BOLD response or a negative percentage signal change, which corresponds to a reduced activation or a deactivation, depending on context/contrast. If your rows correspond to different subjects, then there's one subject who shows reduced activations in the 2nd ROI. That's it. There might be various reasons, e.g. due to anatomical or functional variability the activation peak for that subject might fall outside of the ROI, alternatively that subject might just not recruit this area at all. See it as variability between subjects. In general it's no reason for subject exclusion. Values above 1 just mean the beta/contrast estimates or the percentage signal change is larger than 1. The range is not resticted to 0 - 1 (or -1 to 1). Again, in general this is highly variable between subjects.
Of course you can check the data quality (e.g. look at the mean EPI volume, is there signal loss in that particular region due to susceptibility artefacts, check the motion parameters and their correlation with particular conditions), but you should do so for all of your subjects, as any value, whether positive or negative, could trace back to some artefact.
Hope this helps,
Helmut
|