Investigate the difference between the wPLI code we are using and the other we were using previously

[yacineMahdid]

We are having some issue with the analysis and it might be due by the wPLI generation part. I need to create an experiment where we can see the difference between the two approach given the same dataset.

Attempt 1

When we are using the weighted phase lag index we get (with a step size of 1 seconds): Unconscious:

• AEC = 81%
• wPLI = 69%

Pre-ROC:

• AEC = 68%
• wPLI = 68%

Attempt 2

When we are using PLI only (without surrogate correction and with a step size of 5 seconds) Unconscious

• AEC = 85%
• PLI = 68%

Pre-ROC:

• AEC = 69%
• PLI = 68%

Attempt 3

When we are using PLI only (without surrogate correction and with a step size of 1 seconds) Unconscious

• AEC = 81%
• PLI = 65%

Pre-ROC:

• AEC = 68%
• PLI = 68%

Attempt 4

Redo the analysis with Jason matrices (still looking for how they were generated) Unconscious

• AEC = 85%
• PLI = 78%

Pre-ROC:

• AEC = 68%
• PLI = 78%

Attempt 5

Redo the analysis with the code used by Jason to generate the matrices

• AEC = 85%
• PLI = 81%

Pre-ROC:

• AEC = 68%
• PLI = 76%

Attempt 6

Redo the analysis with Lucrezia code + surrogate correction

• AEC = TODO
• PLI = TODO

Pre-ROC:

• AEC = TODO
• PLI = TODO

[yacineMahdid]

I've found the source of the difference, we just now need to fix it. I'll do a series of bug fix in order to get it right and document it.

[yacineMahdid]

The source of the discrepancy is that Jason used PLI + surrogate analysis with wPLI. Which doesn't hold very well theoretically.