I am performing a source analysis with EEG data. I am dealing now with statistics steps at the source level. Preceding steps include, the averaged conditions (each with the corresponded computation source) within the intra-subject file. At the end of the day I have 14 participants with eight conditions ready to perform the paired t-test. The problem I observed is that, once I run the t-test (with abs(A)-abs(b) option checked), I inspect the statistic map and first, the scale is empty (no values) and second the cortex obviously is empty too. The p-value threshold is set to 0.05, so what I do is to custom the color map and edit the range values mininum=0 and maximum =1, so by this mean what I guess I am doing is adjusting the scale values to the p-value threshold. I don´t understand why still the cortex map is empty. One thing I tried is to check the uncorrected option, and then Brainstorm returns the cortex with colored areas. I run the t-test for all conditions and for all of them, the same problem persist.
The aimed-output accounts for a p-value cortex map, controlling for signals and time and see which scouts I selected previously, are significant in a specific time-window.
Seem that I am missing something here; as far as I know is recommended to select Bonferroni or FDR to make sure you are testing your hypothesis according to the canonical statistics. I would really appreciate some help or suggestion.
What you are supposed to see on the cortex are the values of the t-statistics, but only where the values of p are below the threshold indicated in the Stat tab (corrected for multiple comparisons).
The t values are not bound between 0 and 1, you should not have to edit manually the scale of the colormap for displaying statistical maps.
If you don’t see anything on the brain plus an invalid scale of the color bar, it’s because you have nothing significant at this time point.
Note that the t-tests are computed separately for each time point, you have to move in time to the instants of interest, if you stay in the baseline you will never see anything significant.
To understand how this works, you can little by little increase the p-value threshold in the “uncorrected” mode and you should start to see some colored blobs on the brain, together with a correct scale of the color bar. What you see then is significant with the condition (p < threshold).
If you want to apply a more conservative correction for multiple comparisons, select the options “FDR” or “Bonferroni” in the Stat tab. You can read the corresponding corrected p threshold in the Matlab command window.
The thing is that I don´t see anything when I move over the temporal segment which is from -200 to 1800ms. It could be as you say that is because nothing significant is there but it look kind of weird to me; I have four conditions and a long lasting temporal window. However, this only happens when I apply Bonferroni or FDR corrections but not with uncorrected. I guess I will have to assume that nothing significant is there in the whole dataset but before I wanna make sure that I perform the analysis correctly at previous steps…if you have any suggestion or advice about the procedure that I should check to ensure… that will be very helpful.
I think that if you have something that is shown and makes sense in the uncorrected version, it means that your results are correctly calculated, there is no error before.
The main thing to do is to check the value of the corrected p-value in the Matlab command window when you select “FDR” or “Bonferroni”.
You can recalculate manually the values of the corrected p-threshold for the Bonferroni correction, it will help you understand why the corrected version is so conservative.