1
$\begingroup$

I apologize if this question has been asked previously.

I conduct mixed-effects model analyses in R and MATLAB, so this question is more conceptual than technical. MATLAB and R both produce p-values for each coefficient for GLMERS/fitglme.

In some cases, I may have a model in which there is a 3-way interaction in my fixed-effects structure with just a random intercept term. Here, most of my fixed effects are significant (p < .05).

However, when I add random slopes to the random effects structure, the significance of the fixed effects diminish. I'm assuming this has to do with the random effects accounting for individual variation, but I was curious if there was a fuller answer.

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ It's going to be helpful if you actually state the model and which variables are interacted with which other ones, which are random, etc. Otherwise it's a bit hard to say anything concrete. $\endgroup$
    – conjugateprior
    Dec 6, 2016 at 22:58

1 Answer 1

Reset to default
1
$\begingroup$

Adding random effects to a model leads to shrinkage in fixed effects differences because the random effect variance component is included in the interval of the test. The authors of R mixed procedures do not believe in calculating standard errors and t-statistics for fixed effects as provided in SAS. Instead they use estimates of z statistics to obtain a confidence interval.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.