[R] lme: extract variance estimate

Douglas Bates bates at stat.wisc.edu
Fri Jul 9 03:18:13 CEST 2004

I'm actually replying to Stephen's message but I had already deleted it 
before I read Spencer's message.  (I'm on a slow connection and I am 
simultaneously downloading a big file so I am trying to optimize 
bandwidth, sometimes with unfortunate results.)

In the previous message the author showed the slots u at scale (a numeric 
scalar), u at reSumry (a list of corrmatrix objects), and u at useScale 
(logical).  The quantity that you are looking for is calculated as

  unlist(lapply(u at reSumry, function(x, scal) scal*x at stdDev,
          scal = ifelse(u at useScale, u at scale, 1.)))

plus-or-minus a parenthesis or two.

The reSumry contains the relative variance-covariance matrices in the 
form of correlations and standard deviations.  Some models, such as lme 
models, use a scale parameter.  Others, such as binomial generalized 
linear mixed models, do not.  That is why the estimate of the scale 
parameter is stored separately along with a flag indicating whether 
scaling should be used.

More information about the R-help mailing list