Thanks, Spencer, for spending time on my question.
On Dec 5, 2006, at 1:48 AM, Spencer Graves wrote:
> No, your example computation does NOT produce the desired "lme
> prediction intervals". I ran your script and got exactly the same
> numbers for upper and lower limits. Even without any consideration
> of statistical theory, this suggests either shockingly precise
> statistical estimation or a problem with your algorithm.
There may have been a typo in what I sent to the list.
The following script produces prediction intervals that are not
nearly as shocking as intervals with 0 width:
library(mlmRev)
library(nlme)
hsb.lme <- lme(mAch ~ minrty * sector, random = ~ 1 | cses, Hsb82)
(hsb.int <- intervals(hsb.lme))
hsb.new <- data.frame(
minrty = rep(c('No', 'Yes'), 2),
sector = rep(c('Public', 'Catholic'), each = 2))
cbind(hsb.new, predict(hsb.lme, hsb.new, level = 0))
cbind(hsb.new, rbind(
hsb.int[[1]][1,],
hsb.int[[1]][1,] + hsb.int[[1]][2,],
hsb.int[[1]][1,] + hsb.int[[1]][3,],
hsb.int[[1]][1,] + hsb.int[[1]][2,] + hsb.int[[1]][3,] + hsb.int
[[1]][4,]
))
The latter produces:
minrty sector lower est. upper
1 No Public 12.467015 12.712155 12.95730
2 Yes Public 6.657422 7.387386 8.11735
3 No Catholic 14.522224 15.123113 15.72400
4 Yes Catholic 10.062856 11.821924 13.58099
<snip>
> To briefly outline some of the difficulties, we first should
> ask if you want confidence intervals for the MEAN of future values
> or for the future values themselves?
I want the former, and that's what I think I computed.
> And how do you want to handle the random effects? If you want the
> mean of the fixed effects, averaging over random effects, that
> should be fairly easy.
Exactly. And my question is --- if the CIs I obtained were correct
--- is there a less laborious way to obtain them?
_____________________________
Professor Michael Kubovy
University of Virginia
Department of Psychology
USPS: P.O.Box 400400 Charlottesville, VA 22904-4400
Parcels: Room 102 Gilmer Hall
McCormick Road Charlottesville, VA 22903
Office: B011 +1-434-982-4729
Lab: B019 +1-434-982-4751
Fax: +1-434-982-4766
WWW: http://www.people.virginia.edu/~mk9y/
[[alternative HTML version deleted]]
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
