[R-sig-ME] Seeming discrepancy between summary and confint; was: Confidence interval for relative contribution of random effect variance

[lorenz.gygax at agroscope.admin.ch]

Dear Emmanuel,

Thank you! Stupid me ...

A nice weekend to all of you! Lorenz



> Am 12.09.2014 um 17:55 schrieb "Emmanuel Curis" <emmanuel.curis at parisdescartes.fr>:
> 
> Double check your results, you will see that there is agreement also
> for random effects: the column to use is Std. Dev. which is indeed in
> the confidence intervals given by confint --- just like standard
> deviation for the residuals.
> 
> It just happen that confidence intervals are so wide, that they also
> include the Variance value, but thats « bad luck ».
> 
> On Fri, Sep 12, 2014 at 02:52:11PM +0000, lorenz.gygax at agroscope.admin.ch wrote:
> « Dear Martin,
> « 
> « Many thanks for this explanation which, of course, is very reasonable ;-)
> « 
> « But - and I may be real slow on this - why is the same seemingly not true for the random effects as well (summary and confint give the same absolute values)?
> « 
> « Cheers, Lorenz
> 
> « >> If I do the summary () this is what I get for the random effects part of the output.
> « > 
> « >> Random effects:
> « >> Groups        Name        Variance Std.Dev.
> « >> val:(part:ID) (Intercept) 0.4599   0.6782  
> « >> part:ID       (Intercept) 0.1773   0.4211  
> « >> ID            (Intercept) 0.1278   0.3575  
> « >> Residual                  9.4302   3.0709  
> « >> Number of obs: 1833, groups:  val:(part:ID), 214; part:ID, 72; ID, 25:
> « > 
> « > 
> « >> If I do
> « > 
> « >> confint (HHbT.fin.lmer, method= 'profile')
> « > 
> « >> I get
> « > 
> « >> 2.5 %     97.5 %
> « >> .sig01       0.41713241  0.9210729
> « >> .sig02       0.00000000  0.7535615
> « >> .sig03       0.00000000  0.6697109
> « >> .sigma       2.96898087  3.1786606
> « > 
> « >> Where the above listed variances for the random effects fit nicely into the confidence intervals (.sig0x) but not the value for the residuals / .sigma where the variance from the summary seems to be approximately squared in respect to the confidence interval.
> 
> -- 
>                                Emmanuel CURIS
>                                emmanuel.curis at parisdescartes.fr
> 
> Page WWW: http://emmanuel.curis.online.fr/index.html