[Eeglablist] Regarding .fdt files

[Delorme, Arnaud]

>  - Could I clarify that computing average re-reference always result in n-1 components after ICA?

This is true in general although there can be exceptions.
For example, if you include the original reference (say for example Cz which was 0 as a reference) in the average reference calculation, but remove Cz after calculating average reference, then the rank is the same as the number of channels.

> I also came across on Makoto's pipeline (https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline) that: "When you apply average reference, you need to put the reference channel back if it is missing (which is usually the case). You can include the reference channel in computing average reference by selecting the 'Add current reference back to the data' option in the re-reference GUI. If you do not have initial reference channel location, in theory you should generate a zero-filled channel and concatenate it to your EEG data matrix as an extra channel (which makes the data rank-deficient) and perform average reference. To make the data full-ranked again, one may (a) drop a channel--choosing the reference channel is reasonable (b) reduce the dimensionality by using a PCA option for ICA.”

Yes, this is exactly what I was mentioning above with Cz.

> I do not have initial reference channel

It is 0. You need to look at your montage. To make Makoto’s statement clearer: I would personally only include the reference if it is a scalp channel (Cz, TP10, etc…) not if it is linked ears, nose (rarely used anymore) etc… This is because the average reference assumption is that the sum of all scalp channels if 0 (if uniformly distributed which is never the case - but this is what we do anyway).

Note that it is generally considered that the REST reference is a better choice than average reference, and there is a new plugin in EEGLAB allowing to calculate the REST reference by Li Dong from UESTC, Chengdu, China (it is in the process of being included in the plugin manager and should be there next week).

Best wishes,

Arno