Dear Brainstorm community,
I´m calculating the connectivity A to B between two clusters (frontal and parietal) in one hemisphere. I have a study with many subjects and for some subjects the logfile gives back:
No significant connections were found in this file. Then no connectivity file is created. The problem is now when I try to calculate a paired T-test for the connectivity differences in a pre to post intervention comparison, I have no file to use for this condition for the statistics. Do you have an idea how I can deal with the problem?
Also despite there is a statistic file generated, the logfile reports: Statistical tests on connectivity results are not supported yet. Is it save to use the statistic function on connectivity data?
Thanks,
Boris
Hi Boris,
If you get this message “No significant connections were found in this file”, it is probably because you don’t have enough data in input.
To get a meaningful estimation of any connectivity measure, you need relatively long time series. When there are not enough time samples, the estimated values are replaced with zeros.
To determine what is the smallest time window you can safely use, the only information available is in the code itself (bst_corrn.m, bst_cohn.m).
I don’t know what statistical tests should be used for testing connectivity measures. The parametric and non-parametric processes would run on your files, but it doesn’t mean that the test is adapted to the input data. The selection of the test is your responsibility, we don’t have any recommendations yet.
Our colleagues at USC are working on the methods and documentation of these connectivity measures and should come up with recommendations for the statistical thresholding of the connectivity matrices before the end of the year.
Cheers,
Francois
Hi Boris,
Thanks for your message.
I don’t know which function and measure you are using for computing brain connectivity. Currently, some functions have built-in statistical analysis to only keep significant values, which unfortunately you cannot disable it. Your problem might happen if nothing remain after this analysis as significant connectivity.
We are designing new tools for computing connectivity measures. In that case, you can select whether or not you want to have an analysis for keeping significant values.
So tell me more details and then I can help you better.
Hossein
[QUOTE=Boris Rauchmann;10963]Dear Brainstorm community,
I´m calculating the connectivity A to B between two clusters (frontal and parietal) in one hemisphere. I have a study with many subjects and for some subjects the logfile gives back:
No significant connections were found in this file. Then no connectivity file is created. The problem is now when I try to calculate a paired T-test for the connectivity differences in a pre to post intervention comparison, I have no file to use for this condition for the statistics. Do you have an idea how I can deal with the problem?
Also despite there is a statistic file generated, the logfile reports: Statistical tests on connectivity results are not supported yet. Is it save to use the statistic function on connectivity data?
Thanks,
Boris[/QUOTE]
1 Like
Dear Francois,
I would like to statistically compare the PLV matrices from two conditions (e.g. 15 subject’s matrices for each condition).
Is there a way to do it with the current version of Brainstorm?
Regards
Hello
You should be able to use the Process2 tab to select 15 PLV files on the left, and 15 PLV files on the right.
Francois
Dear Francois,
I get the following warning after processing those PLV files on the Process2 tab.
" warning
15 files […]
Statistical tests on connectivity results are not supported yet."
Sebastian
You can ignore this warning, this is actually functional now.
I removed it from the code. You won’t see it anymore the next time you update Brainstorm.
Hello @Boris_Rauchmann,
Could you explain how you managed to calculate the connectivity on clusters. I am struggling with the procedure. @Francois does it exist in Brainstorm?
I have created a few clusters containing signals from different electrodes contacts (each cluster = one brain region, eg hippocampus, frontal lobe and fusiform gyrus). I would like to calculate the connectivity matrix between these clusters.
Thank you for your help,
Anaïs
You mean clusters of sensors, as defined this tutorial?
https://neuroimage.usc.edu/brainstorm/Tutorials/ChannelClusters
After defining your clusters, you can run the process "Extract > Clusters time series" on your recordings.
This will create new "matrix" files with the signals averaged across sensors.
You can put this files in the Process1 or Process2 tabs to run connectivity analyses on them.