Dear Francois,
I performed a parametric t-test (independent):
To run this test, I selected in the Process2 tab all the deviant trials (Files A) and all the standard trials (Files B). What I obtained is this:
As you probably will remember, my study is on MMN only with healthy subjects (at the moment my goal is simply to confirm what literature claims is true also in our case, so we hope to see negative values in the supra-temporal cortex). For this reason I tried to show the topography (selecting the tame range 100-250ms) either selecting my clusters or not and the result I see is always the same, it doesn't change if I change the cluster. So, I thought to select the cluster before running the t-test (so I wrote the channel names in the "Sensor types or names" field in the pipeline editor). But now I have the problem that it is impossible to show a topography. I tried everything I knew but every time I see the same result:
This is totally discordant from what I see in the t-test graph with the selected cluster. This cluster is negative (in the 123-239 ms latency time I selected), so I should see only a blue area in the topography chart. This is not what I see, for this reason I think that selecting the cluster is not relevant for what shown in the topography.
Can you suggest me something to solve this problem? What I expect to see is only a blue area corresponding to the negative t-value I should have.
Thank you so much,
Gabriele
I’m not sure I clearly understand what you did.
If you’d like me to have a look at your results, please send me an example dataset:
- Duplicate the subject in which you stat file is: right-click on the subject folder > File > Duplicate subject
- In the new subject: keep all the anatomy files, the channel file, the stat results. Delete all the other files, not necessary to replicate your unexpected observation.
- Right-click on the subject folder > File > Export subject
- Upload the zip file somewhere (dropbox, google drive, etc…) and post the download link here
Thanks
Francois
This is the link:
https://www.dropbox.com/sh/0i9uq8hhwzxglwu/AADXQPmBn-2xykx0d_KSLnDVa?dl=0
I have uploaded 2 zip files (one of them also has the "1" or standard, and "2" or deviant trials; I decided to keep them in this folder just in case you want to do some operations on them). The second zip file has only the channels, the anatomy and the stat.
I can't see on the topography the t-test performed on the selected cluster. I need something to show only the t-test in the left-frontal and right-frontal cluster. More precisely, what I expect to see is negative t-values in my clusters (because of the MMN principles). I can see that these clusters have negative values but when I show the topography data are different (as you can see in the previous pictures).
Thanks again
Gabriele
I think there is nothing wrong with the files you sent me, but your interpretation of what negative values mean seems incorrect.
In MEG topographies, just above an active brain source, you observe a pattern with high negative and positive values. When you test a topography A against a topography B with the same pattern but with lower amplitudes everywhere (t-test A vs B), you observe both positive and negative values, while A is “stronger” everywhere. Conclusion: unlike EEG, the sign of MEG signals (and of the difference of signals) is ambiguous. To get around this issue, test the absolute values or power.
Read carefully the introduction tutorials, I think this problem is properly explained:
https://neuroimage.usc.edu/brainstorm/Tutorials/Difference
https://neuroimage.usc.edu/brainstorm/Tutorials/Statistics
https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows
For general introductions to MEG signal processing, we recommend some readings at the bottom of the tutorials page:
https://neuroimage.usc.edu/brainstorm/Tutorials#Background_readings
