I have read that the distribution for the random effects estimator in lme4 is highly skewed and for this reason standard errors are not reported. I wonder if anyone can provide a reference for this ? I have access to the book by Bates and Pinherio, but not Raudenbush and Bryk. A published paper would also be fine. Thanks !

  • $\begingroup$ Not exactly what you want, but see bottom of p18 here for some more details: lme4.r-forge.r-project.org/book/Ch1.pdf $\endgroup$ – P Sellaz Feb 18 '12 at 16:03
  • $\begingroup$ To clarify, by 'distribution of random effects estimator' do you mean the distribution of the variance of the random effects? $\endgroup$ – onestop Feb 18 '12 at 20:39
  • $\begingroup$ Yes, that's what I mean. $\endgroup$ – Robert Long Feb 18 '12 at 23:32

I don't know of anything offhand, aside from Doug Bates' books and his postings on various online forums. That should be sufficient to justify why you don't report them. But if you want to try and quantify uncertainties, I might try a Bayesian approach, i.e., simulating from posterior distributions of variance parameters using MCMC. Maybe look at:


  • 1
    $\begingroup$ Welcome to the site, Cam. I sure hope to see you here as often as I see you on SEMNET, multilevel and statalist :) $\endgroup$ – StasK Feb 18 '12 at 17:49
  • $\begingroup$ Am I right in thinking that's what is going on here: example(mcmcsamp);densityplot(samp0);qqmath(samp0) ? $\endgroup$ – P Sellaz Feb 18 '12 at 18:17

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.