Announcement

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

  • Interpretation of Intraclass Correlation after Multilevel Model

    I have used the postestimation command "estat icc" to estimate intracorrelation coefficients after two multilevel models (command "mixed", random effects at one level; ~10k observations each). I am now looking to interpret my results (.133 in one case, .182 in another). Would you consider these results high or low when it comes to an ICC?

    My ultimate goal is to determine whether a multilevel approach is necessary, or whether I should use GEE instead.

    Many thanks!

  • #2
    Well, if you Google rule of thumb magnitude of ICC and look at what it brings up, then it would seem that values in the range of 0.1 to 0.2 are not considered very impressive by those in the know.

    I'm curious how that plays into the decision to choose between a population-average approach (GEE) and an individual-specific approach.

    Comment


    • #3
      Thank you for your response! I was torn between both approaches, as MLM seemed more fitting intuitively, and GEE conceptionally. I've looked at more arguments in the meantime, and decided to use GEE.

      Comment

      Working...
      X