[R-sig-ME] SEs for prediction?

[Joshua Wiley]

Hi All,

I have some LMMs fit with lme  (EEG measures on different brain
regions which require a non independent residual covariance structure,
on kids, nested within families).  The random effects are just
nuisance effects in this case.  The study PI would like adjusted means
for different conditions and standard errors.

Now I know that there is not a simple way of calculating accurate SEs
for prediction, but before I roll my own code to draw from the
posterior, bootstrap, or just move the whole analysis to a Bayesian
framework, I wondered if others had crossed this bridge and had a any
suggestions for a quick and dirty approach.  The SEs are just being
used in presentation with the adjusted means, not really the model or
inference, so I am not terribly concerned about them being optimal.

Thanks as always for input,

Josh


--
Joshua Wiley
Ph.D. Student, Health Psychology
Programmer Analyst II, Statistical Consulting Group
University of California, Los Angeles
https://joshuawiley.com/