1

This is a very simple question, but I'm starting to doubt myself:

I have a mixed model which I estimate in Matlab with fitglm

mdl = fitglme(data, 'DV ~ 1 + IV + (1+IV|subject)')

The following piece of code extracts the random effects coefficients (i.e. the individual b0 and b1's)

[B,BNames,stats]= randomEffects(mdl)

Now I'm wondering: Are those resulting coefficients deviations from the main fixed effect coefficient? So, say to interpret an individual random intercept, I would need to take the main intercept (fixed) and add the random intercept to it? And the same with the random slope?

Maria
  • 405
  • 2
  • 6

1 Answers1

1

There's no generally correct answer to this, as it depends on the specific software. Even within the same mixed-modeling software, whether you get deviations from the fixed effect or predictions including the fixed effect can depend on just how you interrogate the model; see this page for an example.

The best way to tell (if the manual page isn't clear) is to examine the distributions of the values of the random effects. If they center around 0 (for a corresponding non-zero fixed-effect value) then they are deviations from the fixed-effect estimates.

EdM
  • 57,766
  • 7
  • 66
  • 187