Dear list, Could anyone tell me what is the scale of the variance components (tempL.x parameters) in the mcmc output from glmmadmb (version 0.7.7)? From previous posts it seems that they are supposed to be log of the standard deviation in the random effects distribution, but I don't get any negative values (I expect variance <1) so I guess that has changed. Are they variances or standard deviations? I presume the variances/standard deviations are on the scale of the linear predictor. Cheers, Torbj?rn Ergon
glmmadmb mcmc - scale of variance components
2 messages · Torbjørn Ergon, Ben Bolker
9 days later
On 14-01-15 05:18 AM, Torbj?rn Ergon wrote:
Dear list, Could anyone tell me what is the scale of the variance components (tempL.x parameters) in the mcmc output from glmmadmb (version 0.7.7)? From previous posts it seems that they are supposed to be log of the standard deviation in the random effects distribution, but I don't get any negative values (I expect variance <1) so I guess that has changed. Are they variances or standard deviations? I presume the variances/standard deviations are on the scale of the linear predictor.
The vignette (vignette("glmmADMB",package="glmmADMB")) says (p. 15)
that tmpL contains the RE variances on the standard-deviation scale.
(And yes, this is on the scale of the linear predictor.)
(I'm not claiming this information is easy to find!)
cheers
Ben Bolker