Questions about decoding

Discussions
1

Dear Brainstorm team,

I have performed the decoding procedure with SVM, and obtained interesting results when comparing 2 classes (>90%) in most of the control subjects. Thank you for this excellent feature.

These are my questions:

  1. Would be possible to get the temporal curve (average decoding accuracy) plus the standard errors across trials for ONE subject? I am only able to compute the std errors when averaging all the subjects.

  2. What is the optimal number of electrodes to run MVPA?

  3. I don't think is implemented yet, but is it possible in the near future to select the electrodes at the most informative sites? This could reduce noise. I am typically using 64 channels.

  4. How can I implement statistical analysis on the decoding results?

Thank you very much in advance

2

Would be possible to get the temporal curve (average decoding accuracy) plus the standard errors across trials for ONE subject?

The standard error you can display on the average is the standard error of the average itself... this information doesn't exist on a single file.

For all the other questions:
@pantazis @Sylvain @MartinC ?

3

Ping?