R help - Dec 2010 - [R-lme] Extract estimated variances from output of lme?

Hi all,

I have the output of summary() of an lme object called "lme.exp1", for
example

#############################################
> summary(lme.exp1)
Linear mixed-effects model fit by REML
  Data: DATA
  Log-restricted-likelihood: -430.8981
  Fixed: fixed.exp1
    ....

Random effects:
 Formula: ~-1 + mu1 + log.sig1 | animID
 Structure: Diagonal
             mu1  log.sig1 Residual
StdDev: 2.605223 0.1013134 3.876431

Correlation Structure: General
 Formula: ~1 | animID/fieldN
 Parameter estimate(s):
 Correlation:
  1
2 0.612
Variance function:
 Structure: Different standard deviations per stratum
 Formula: ~1 | type
 Parameter estimates:
       mu1   log.sig1
1.00000000 0.03722675
Number of Observations: 376
Number of Groups: 32
##############################################

If I want to reuse these estimates, I can use
> summary(lme.exp1)$coefficients$fixed;
to extract fixed effect estimates, and

 > summary(lme.exp1)$sigma

for the common variance parameter sigma.   But if I need the covariance
estimates 0.612 and 0.0372, do I have a function to extract these numbers
too?   I'm looking for something similar to function ranef() [of course
ranef() does not serve my purpose].  Any advice appreciated!

Tingting

	[[alternative HTML version deleted]]

Tingting Zhan <tingting.zhan <at> jefferson.edu> writes:
> 
> Hi all,
> 
> I have the output of summary() of an lme object called
"lme.exp1", for
> example
> 
> #############################################
> > summary(lme.exp1)
>
  [snip]
> for the common variance parameter sigma.   But if I need the covariance
> estimates 0.612 and 0.0372, do I have a function to extract these numbers
> too?   I'm looking for something similar to function ranef() [of course
> ranef() does not serve my purpose].  Any advice appreciated!
> 
> Tingting

  ?VarCorr

and

str(VarCorr(lme.exp1))

to figure out how to extract the values you want (it's
a slightly odd object, a matrix of mode 'character' with
labels interspersed with numeric values ...)

  r-sig-mixed-models is probably the best list for this
sort of query ...