[R-sig-ME] Yet another question on CIs for coef(model) extracted conditional coefficients

[Ben Bolker]

On 13-12-18 05:46 PM, Richard Asturia wrote:
> Dear list, this one is probably already answered somewhere, but to the best
> of my efforts I did not find precisely the information I am looking for,
> although I did find a lot of useful comments about the overall subject.
> 
> So, what I wanna do is the usual procedure of calculating CIs for the
> coefficients at the grouping levels, i.e. the ones extracted with
> coef(model) , which are the same as ranef(model) + fixef(model).
> 
> I found the explanation on the FAQ on how to extract the standard errors
> for the conditional means with ranef(model, condVar = TRUE), but I still
> don't know what would be the best practice here for using those to
> calculate CIs. The FAQ does not seem to discuss that particular issue.
> 
> I know that functions such as dotplot and coefplot2 give me intervals, but
> seems to me as a rather naïve way of calculating those. Please, is there
> any suggestions you could make on how to accomplish that, i.e. how to
> calculate CIs for the conditional coef(model) coefficients, by using the
> standard errors extracted with condVar=TRUE?

  This is not easy.  The following (long) thread is probably the best
discussion I've seen:

http://thread.gmane.org/gmane.comp.lang.r.general/286283/focus=9589