source_roi_power and event-related data

[nguyen-td]

We received some questions via email, do you think you can provide some answers? @arnodelorme

1. What is the scale of the output in EEG.roi.source_roi_power? Our values are quite large (see attached excel sheet – this is averaged over the theta band but the large numbers are present for all frequencies). The MIM values however seem to be in line with what is expected (i.e. between 0 and 1).
2. We realise that the package was initially for resting state data but can now be used on event-related data. Is there a setting we need to change or any different functions we need to use to run the package on event-related instead of resting state data?
3. Similar to this point, at what stage does the function extract trial information? We were having an issue with trial numbers but could not identify where the trial information was being read in.

I can redirect the excel sheet in 1) via email if necessary.

EDIT: I just received the permission to upload the data: sampledataoutput.xlsx

[arnodelorme]

You can compress the Excel file in a zip and attach to this thread if needed.

1. I think this is a question for @stefanhaufe
2. It cannot yet be used for evet-related data. The function extract 2-second data segments (called trials) but does not process the time information.
3. See (2).

[montanahunter]

Thanks so much for your reply!
Since the function does not process time information, we are now planning to pre-select time windows. We will average over these windows before running the function. Would this approach be appropriate?

[sahin-ozsoy]

I am using ROI Connect on epoched data, resampling to 100 Hz, and extracting 2 second epochs locked on stimuli. I believe if you are expecting a consistent information flow across the 2-second epoch across trials, the MIM or TRGC can still generate estimates of such FC. If you are interested in event dynamics, you can always use ERSP, ITC, and cross-coherence measures.

[montanahunter]

Hi

Just following up on this again. We are trying to run the ROI connect package on event data but selecting a smaller window of time (200ms). However, this seems to be throwing off the epoch and event numbers. I am also wondering if there is a concern about running this code on only 200ms of data. Any help or advice with this would be greatly appreciated!