Skip to content
Prev 9599 / 20628 Next

lmer - BLUP prediction intervals

Hi Harold,

I think that we're on the same page.  My opinion is that the estimates
and the predictions are both random variables.  They have a joint
distribution under the model, and it is possible to estimate the
parameters of that joint distribution, under the model.  

I tend to think that if I can't write down their joint distribution
then I'm not sure what are the grounds for taking any linear
combination - including the very useful adding of them - either.

If obtaining the covariance term is reasonable within epsilon time and
effort, then I would advocate doing that and following your
prescription.

Andrew.
On Thu, Feb 07, 2013 at 09:45:38AM -0500, Doran, Harold wrote: