[R-sig-ME] Extracting Information from nlme ( )

[Ben Bolker]

On 02/18/2011 04:13 PM, Jeffrey Harring wrote:
> Question...
> 
> Is there a function or nlme object that allows one to extract the
> standard errors of the variance/covariance parameters within nlme ( ).
> For example, I would like to extract the SE's for the
> variance/covariance matrix of the random effects (which is a 2 by 2). I
> can get the actual parameters, but I am having difficulty extracting
> these parameters SE's.
> Any suggestions would be welcomed.
> 
> Thank you,
> Jeff
> 

  me again ...

?intervals.lme
  intervals(nlmefit,which="var-cov")

 You can also extract the raw information via
nlmefit$apVar

 In his current incarnation, Doug Bates would tell you that the sampling
distributions of the variance-covariance parameters are so non-Gaussian
that basing inferences on their approximate standard errors is dangerous.

  cheers
    Ben Bolker