LIMO-EEG-Toolbox/limo_tools

Summary of discussion 7/19

Opened this issue · 7 comments

To do LIMO

  • Think about course plot and how to make this interface better
  • Fix contrast manager issue.

To do EEGLAB (pop_limo and std_limo)

  • add full factorial interaction in case there is a single subject (add this as an option)
  • full factorial - never do across subject (cannot do ANOVA across subject - no parameter interaction)
  • removed bootstrap from plotting LIMO interface 48fb22e
  • Fixed precompute fd890af

problem to solve: when precompute boostrap --> plot --> the figure GUI cannot find the H0 file?? yet, reloading the figure makes it work -- ie path issue to fix when calling clustering

Fix preloading LIMO expected channel --- I recommend not to
This file should really be looked at as the automated stuff doesn't always do a good job ; if using STUDY it's there

course plot -- fixed the GUI, @arnodelorme to play with limo_central_tendency_and_ci and see how to improve limo_add_plot

add full factorial interaction --> won't be a simple fix because all we have are conditions, and conditions can be pulled in many different way to create factor -- we could make assumptions that event types pooled together (for which we compute contrasts) make factors and create new factor variables (there is no need to build the interaction, just pass the flag and limo_design_matrix builds it)

example: user selects famous, unfamiliar, scramble faces as variable 1 and repetition 1,2,3 as variable 2
--> currently still make 9 conditions but compute contrasts for faces and repetitions
--> the proposed solution would be to create a variable face (single column with 1,2,3) and variable repetition (single column 1,2,3), these flag to trials, and send that to limo_batch with the option full factorial on (and it will deal making the 3+3+9 columns)

To do LIMO

Think about course plot and how to make this interface better
Fix contrast manager issue.
To do EEGLAB (pop_limo and std_limo)

add full factorial interaction in case there is a single subject (add this as an option)
full factorial - never do across subject (cannot do ANOVA across subject - no parameter interaction)