Forum Replies Created

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • in reply to: Mixed model specification and centering of predictor #346
    SeeArrr
    Participant

    Thanks for the explanation. Helped a lot. I decided to go for the m2 with interaction in the random slopes.

    As a next step I would like to introduce time as another IV. I have the baseline EEG resting state recording (time 0) and a post treatment EEG recording (time 1) as well as pre- and post-distress scores. I wonder, due to the fact that I just have two time points, that distress might be a mediating variable. I haven’t found anything about mediation in the documentation, so I wonder if there is a recommended package by you Henrik?

    Thanks again!!

    in reply to: Mixed model specification and centering of predictor #342
    SeeArrr
    Participant

    Thank you for your fast reply! At the same time, my hopes are crushed.

    I’m working with a disease that causes alternation in the resting state EEG pattern. In the literature, however, it has not yet been possible to make a clear statement as to which alternation takes place in which brain region. In addition, the experience of the disease can alter a lot between individuals, which results in quiet a heterogeneous EEG pattern. Therefore, I think the approach of using not only the patient but also the electrodes as random effects might be more appropriate. But I haven’t found anything in the literature yet. In addition, I use a questionnaire which records the distress of this disease (and hopefully is mirrored in the EEG pattern).

    The dependent variable Power contains the EEG measure. I am interested if this measure depends on distress (questionnaire – continuous variables) and FrB (8 frequency bands, factorial), as well as duration of the disease. Random effects are PatID (participants) and Electrodes (65)

    m1 <- mixed(Power ~ FrB * Distress + Duration + (FrB | PatID) + (1 | electrodes), data, method=ā€Sā€)

    As far as I experienced it, other studies are working with electrode clusters (and not single electrodes). This would result in my opinion in:

    m2 <- mixed(Power ~ Cluster * FrB * Distress + Duration + (FrB + Cluster | PatID), data, method=ā€Sā€)

    What do you think?

    in reply to: Mixed model specification and centering of predictor #339
    SeeArrr
    Participant

    I stumbled upon this post with great curiosity. Since I also work with EEG, but under a somewhat different, but nevertheless similar approach, I wondered whether there is a published paper to your computations and considerations?

    @Henrik, please excuse me for using your forum for such a purpose.

    @cbrnr, please excuse my direct request, but your help would be a great support.

    Thank you

    • This reply was modified 1 year, 3 months ago by SeeArrr.
Viewing 3 posts - 1 through 3 (of 3 total)