It seems like in order for me to generate contrast plots between two conditions to range from -1 to 1 (the weights), I can only use one gradiometer (either planar1 or planar2) to avoid the issue of having a ‘positive only’ plot due to the absolute value outcome from combining planar1 and planar2 for ‘grad’ type plots.
When I’m looking at my data however, there are a few pairs of grad sensors that has considerable difference between planar1 and planar2 activity. The below plots show the TFRs of a comparison of parietal sensors 2012 and 2013 of a neuromag machine that has undergone one way anova followed by cluster-wise correction.
Although these two gradiometer sensors are paired up (planar1 and planar2), the outcome is greatly different with one showing significant cluster activation but not the other.
I’d like to ask if :
this is a normal occurrence and
if so, is it methodologically valid for me to hand pick the planar sensor within the pair that produced the stronger activity, and assume that the signal in this one sensor is indicative of the activity within the corresponding region of the brain (despite the other sensor in the pair showing otherwise)?
Appreciate any advice from experienced MEG analyst out there!
Thanks Alex. I see that there is a merge_grad function with plot_topo on evoked data. But there isn’t one when it comes to plot_topomap, which is the ideal plotting method for my analysis.
Is there a way to do it for the latter?
Nevertheless, if it is not possible, what I assume is that taking the RMS of the power sums up the activity across both planes in a given spot. So if I decide to just use one sensor within the pair for statistical analysis, the only setback I would get is reduced power coming from a given location of the brain? And that false positives should not be an issue (it would be if power is calculated by getting the mean of the two sensors)?
If the above is true, then I am perfectly fine with it.
