T test against zero output

[frtalami]

Dear Brainstorm Team,
I kindly need some help on how to correctly use the t test against zero process on my dataset.
I have a big dataset of 1288 SEEG contacts for 100 trials. I obtain this dataset combining together the electrodes from 20 patients and creating a “fake patient” that I imported in brainstorm.
In brainstorm I segmented the data around the event [-200;500ms] and then I did a time-frequency analysis (Morlet) using the brainstorm process, saving individual time-frequency maps for each trial. Then I did a z-score transformation relative to baseline (always taking into account the edge effect), obtaining individual z-score maps for each trial.
What I want to do is to identify the responsive leads comparing the gamma band power in each post-stimulus bin against the zero.
To do so I selected the z-score maps in the Process 1 tab and run the process Test>Parametric Test against zero: I run a One-sample Student t-test and selected “Average selected frequency band” and “Match signals between files using their names”. My goal is to compute a 1 sample t test to determine the time bin significance across population of responsive leads.
This is what I get, which is great, but I have a question about the time series visualization.

How come there is a steep increase (red arrow) with all the time series starting at 5 or something instead of zero? It changes when I play with the significance threshold and the correction for multiple comparison, so I guess it is something related to that, but I cannot understand what it is.
Is there a way to change the visualization or am I missing something?

I would also like to map all the significant leads at any time (if a contact is significant at any time=1 else 0) and the delay between responses following stimulation as the first significant bin of activation. Is there a way to do this in brainstorm? These are things I could do in Matlab, but it would be lovely to use the brainstorm machinery to visualize the results. Could I do that using the script generator?

Thank you very much
Francesca Talami

[Francois]

@frtalami

How come there is a steep increase (red arrow) with all the time series starting at 5 or something instead of zero?

In the time series figures, the non-significant values are represented as zero. If you click on one signal, you can observe that after the sharp step, it keeps going at zero until the end of the epoch. This step you observe is only an artifact of this visualization. You can safely ignore these vertical lines.
What is not visible above is not significant.

I would also like to map all the significant leads at any time (if a contact is significant at any time=1 else 0)

I'm not sure how you could create this result using the interface.
But you can compute this manually and save it as a new data file in your database

• Apply the statistics threshold: https://neuroimage.usc.edu/brainstorm/Tutorials/Statistics#Convert_statistic_results_to_regular_files
• Export the file to Matlab: https://neuroimage.usc.edu/brainstorm/Tutorials/Scripting#Export.2FImport_with_the_database_explorer
• Create a data file with non-zero values if there is significant value at any point in time (duplicate time as data files must have two time points minimum):
  DataMat.F = repmat(double(any(DataMat.F,2)), 1, 2);
DataMat.Time = DataMat.Time([1 end]);
• Right-click on the folder > File > Import from Matlab > DataMat

the delay between responses following stimulation as the first significant bin of activation. Is there a way to do this in brainstorm?

How would you like display these delays? As the value associated to each sensor?
You can probably do something similar as what I described previously.

Have you followed completely the tutorial about epileptogenicity maps? It includes the computation of delay maps, computed in a similar way that what you describe here, but over the anatomy.
https://neuroimage.usc.edu/brainstorm/Tutorials/Epileptogenicity

For help with scripting in Brainstorm:
https://neuroimage.usc.edu/brainstorm/Tutorials/Scripting