Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Estimate effect size for interaction in (xt)mixed

    I would like to estimate the effect size of an interaction term in a linear mixed model.

    The syntax for my model is:

    mixed y a##b || id:, var reml


    Following the idre UCLA web page "How can I estimate effect size for xtmixed?" I came up with the following formula to estimate Cohen's f, though I am not sure it is correct:


    f2a#b =


    R2a##b - R2ab
    ------------
    1 - R2a##b



    Could somebody confirm that this is correct?


    Furthermore, a has two levels and b has 7 levels. Would it be possible to estimate the effect size per interaction level? If so, how.

    Thanks for your help.
    Last edited by Vincent Koppelmans; 18 Dec 2014, 14:22.

  • #2
    In reply to my previous post:

    To solve my problem I have created dummy variables for each interaction.

    (For verification I compared my initial model with the model including a, b, and all interaction terms of a and b. These two models were identical in output)

    By running models in which I left out single interaction terms, and comparing those models to the full model (using above described formula for f2) I estimated the effect size for individual interaction terms:

    f2a1*b1 =


    R2full_model - R2model_without_a1*b1
    -------------------------------
    1 - R
    2full_model

    Comment

    Working...
    X