[Eeglablist] help with group-level interactions between variables in LIMO EEG

Dr Cyril Pernet cyril.pernet at ed.ac.uk
Wed Sep 10 08:50:34 PDT 2014

Hi Jasna,

>I've got a design in which participants categorise three types of  images, which are displayed with three different types of contrast  (luminance and 
> two chromatic types of contrast). LIMO EEG seems to be  ideal for analysing this data, as it allows for continuous variables to  be included in the  
> design. On the single subject level, I have created a design matrix which suits  my study, with 3 levels of my first factor as categorical and then the 
> three continuous variables in addition to that. I get 1 categorical  effect and 3 covariate effects for each subject.
> At group level, I have obtained the effects of the 3 co-variates on the  ERP and they look very sensible, as one would predict it.

ok so I assume here you ran 1c sample t-tests for each covariates (presumably betas 4, 5, 6)

> What I would still like to do is this:
> 1) check for partial correlations between my three covariates, to see  how independent they are of each other, preferrably at the group level.

all I can think about is to compute the partial correlations per subjects and then do some group stats on that (I would need to think about that, you can email me off line if you wish)

> 2) check if the covariates interact with the categorical variable, again preferrably at the group level.

to do that you would need a more complex model per subject, specifically you want a model with 9 covariates that is split the covariates per conditions and add 0  -- importantly (1) zscore the 9 covariates before adding the 0 and (2) tick (or set in the LIMO.mat if you script) the do not zscore before estimating the models

the output is now the effect of each covariate per condition, so you can run a repeated measure ANOVA between covariate of a condition for instance luminance cond 1 vs luminance cond 2 vs luminance cond 3, therefore testing if luminance is changing when changing consditions = interaction


Dr Cyril Pernet,
Centre for Clinical Brain Sciences (CCBS)
Neuroimaging Sciences

The University of Edinburgh
Chancellor's Building
Room GU426D
49 Little France Crescent
Edinburgh   EH16 4SB 

cyril.pernet at ed.ac.uk
tel: 0131 465 9530

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140910/b4a2193a/attachment.html>

More information about the eeglablist mailing list