[Eeglablist] STUDY structure in EEGLABv2021

[Delorme, Arnaud]

Hi Agatha,

> 1) How do we specify a path for design files?

This path was removed after EEGLAB 14 (so in EEGLAB 2019.0). It was most useful when comparing parameters for ERSP or spectrum, for example.

There is no longer a path for STUDY designs when using single trials (in EEGLAB 2020 and later versions). They all share the same files, and files are precomputed only once. The precomputed trial measures are pulled on the fly for plotting and statistics. This allows for more complex statistics (e.g., LIMO integration). Files are larger, but there is no noticeable slow down when working with a solid-state drive.

We could have kept the option to have different sets of precomputed files but:
- We would have had to create an entirely new mechanism to do so and allow people to switch between sets of precomputed values. Only a handful of people used that feature, and it added a lot of complexity to the code (so time involvement plus potential bugs)
- As you could note with EEGLAB 14, the ability to define a custom folder added some instability. For example, when you copied the STUDY, then the manually set path was absolute and pointed to the old study. There was no real way of fixing that issue. There were also other minor issues which you reported.
- An alternative exists. The best alternative is to make a copy of your STUDY and recompute measures.

So to make a long story short, we removed that option to avoid having to reimplement it, to increase stability, and also because an alternative exists.

> 2) How do specify subsets of events to include in the design?

To specify specific subjects, you can set the default at the top of the STUDY design GUI. This will apply to all designs, though. In general, it is not good practice to handpick subjects for specific designs. If you want only to process one group of subjects, you can do so by selecting the group independent variable and the value you are interested in. In that case, only subjects of that group will be included.

> 3) Is there a change to the design files that would preclude them from being recognize in the pop_chanplot() viewer?

Yes, cannot use EEGLAB 14 pre-computed files. These contained averaged measures, and EEGLAB 2019 and later use single trials. You need to recompute these files.

Best wishes,

Arno