 |
 |
Dear Vladimir and Gareth, We performed statistics to compare two experimental conditions on source-reconstructed MEG data (using MNE/IID). The data were reconstructed over the entire epoch (-200 to 600 ms post-stimulus) and windowed (through the SPM12 option following the source reconstruction step) with a sliding time window of 50 ms. Statisical analysis performed on the source reconstructions results windowed over 325-375 ms revealed a significant difference (pFWE < 0.05) between the two conditions in the superior parietal region (peak coordinate = 10 -48 72; see Figure attached left panel). We then extracted the time course of the sources activity for that specific coordinate (x y z) in each condition (right panel), as we wanted to confirm the presence of such difference over time on that specific coordinate (x y z). This was done with the function used by SPM for virtual electrodes (spm_eeg_inv_extract.m). Here is the issue, as you can see in the right panel of the figure attached, we couldn't detect the difference between 325-375 ms at that peak coordinate. Do you have any insight on what may explain such discrepancies between the 2 sides of the same coin? Of course everything (number of subjects, conditions, etc..) was exactly the same. Could this discrepancy be related to different computations of the covariance matrix in the two cases? (computed on the 50 ms time window (325-375 ms) in case 1 (left panel) and computed on the whole time window (-200 - 600 ms) in case 2 (right panel) ? If so, what would be the best solution to observe the time course of our effects? Is there a way to compute the time series but centre the computation of the covariance matrix on the specific time window of interest? Thank you very much for your help. Best wishes, Veronica