[R] bootstrap, lme, random effects

[Andrew Beckerman]

Hi there.  OSX/R2.0

We are trying to implement a bootstrap of the coeffecients of a mixed  
effect model. In particular, we are interested in the intercept and  
slope of the random effects.

Following from the basics for a linear model, we construct our lme  
models and a boot function:

library(nlme)
library(boot)
data<-read.csv("~/data.csv")

bootcoef<-function(data,index){
dat<-data[index,]
mod<-lme(Frames~Man2+Manip+Strings+Date.+Cut.,random=~Man2|ID.,data=dat)
fixef(mod)
}
boot.out<-boot(data,bootcoef,99)
boot.ci(boot.out) # produces information via boot.ci() that suggests  
this is not necessarily successful

ORDINARY NONPARAMETRIC BOOTSTRAP
Call:
boot(data = data, statistic = bootcoef, R = 99)

Bootstrap Statistics :
         original       bias    std. error
t1* 16.125904015  5.299827478  9.98818463
t2*  0.010901682 -0.004134585  0.01621935
t3* -0.038168126 -0.078833467  0.35778286
t4*  1.101486342 -0.021886290  0.45720400
t5*  0.005982241 -0.009140839  0.01563175
t6*  2.729537567  0.287663533  1.56150779
 > boot.ci(boot.out)
BOOTSTRAP CONFIDENCE INTERVAL CALCULATIONS
Based on 99 bootstrap replicates
CALL :
boot.ci(boot.out = boot.out)
Intervals :
Level      Normal              Basic             Studentized
95%   ( -8.75,  30.40 )   ( -8.40,  36.01 )   (-40.93,  35.15 )
Level     Percentile            BCa
95%   (-3.75, 40.65 )   (-5.55, 28.65 )
Calculations and Intervals on Original Scale
Some basic intervals may be unstable
Some studentized intervals may be unstable
Some percentile intervals may be unstable
Warning : BCa Intervals used Extreme Quantiles
Some BCa intervals may be unstable
Warning messages:
1: NaNs produced in: sqrt(tv[, 2])
2: Extreme Order Statistics used as Endpoints in: norm.inter(t,  
adj.alpha)

As stated above, we are interested in the ranef(mod) components.....  
including this instead of fixef(mod) results in the error:
 > bootcoef<-function(data,index){
+ dat<-data[index,]
+  
mod<- 
lme(Frames~Man2+Manip+Strings+Date.+Cut.+ID.*Man2+ID.*Manip,random=~Man2 
|ID.,data=dat)
+ ranef(mod)
+ }
 >
 > boot.out<-boot(data,bootcoef,99)
Error: incorrect number of subscripts on matrix

Suggesting that the setup of the ranef(mod) list is different  
(clearly)....

Any suggestions on any of this?  I have a sneaking suspicion this is  
not a straightforward issue.

Cheers
andrew

------------------------------------------------------------------------ 
---------
Dr. Andrew Beckerman
Department of Animal and Plant Sciences, University of Sheffield,
Alfred Denny Building, Western Bank, Sheffield S10 2TN, UK
ph +44 (0)114 222 0026; fx +44 (0)114 222 0002
http://www.shef.ac.uk/beckslab
------------------------------------------------------------------------ 
----------