calculation of confidence intervals for random slope model
On Mon, Nov 16, 2015 at 5:56 AM, Henry Travers
<henry.travers at zoo.ox.ac.uk> wrote:
I have what I hope is a relatively straightforward question about how to interpret the results of a mixed effects model of the form: fm1 <- lmer(Reaction ~ Days + (Days | Subject)) I am running an experiment such that I am most interested in the (equivalent of the) effect of Days for each Subject, rather than say fitted values. I understand how to derive the point estimates for this effect, but I am struggling to see how to calculate confidence intervals for these estimates that take account of both the standard error in the parameter estimate for Days and the uncertainty in the corresponding slope estimates for each Subject. I would be very grateful if someone could point me in the right direction or to a suitable reference.
This is a surprisingly difficult question to answer. There have been extended discussions in the mailing list (which I don't have time to dig for now) about whether it's OK to add the conditional variances of the conditional modes to the variances of the fixed-effect predictions, and the circumstances under which this would be (in)accurate/(anti)conservative. The other alternative is to use bootMer + predict to get confidence intervals ... This should probably be added to the FAQ ...
------------------------------------------------
Henry Travers, PhD
Research Associate
Interdisciplinary Centre for Conservation Science
Department of Zoology
University of Oxford
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models