Skip to content

lmer model and mcmcsamp

4 messages · Emma Jones, Cyrus Shaoul, Dennis Murphy +1 more

#
Hi:

See
http://glmm.wikidot.com/faq

and pay particular attention to the sections labeled
* Why doesn't lme4 provide p values/denominator degrees of freedom?
What other options do I have?
* Implementations of MCMC and parametric bootstrap

That will give you a fairly broad view of the state of affairs at the
present time.

HTH,
Dennis
On Thu, Aug 11, 2011 at 9:36 AM, Emma Jones <stp08emj at sheffield.ac.uk> wrote:
2 days later
#
Hi,

I have had similar problems - I came to the conclusion that a) I had  
to make some transformation such as log something/divide by sigma etc   
b)  the "locally-uniform" priors are i) not uniform or ii) uniform but  
in the wrong locality or c) mcmcsamp is broken.

It would be good to get some clarity; mcmcsamp even seems to behave  
oddly for the example data set:

fm1 <- lmer(Reaction ~ Days + (1|Subject), sleepstudy)
fm1.mcmc<-mcmcsamp(fm1, n=1000)


summary(fm1)

Linear mixed model fit by REML
Formula: Reaction ~ Days + (1 | Subject)
    Data: sleepstudy
   AIC  BIC logLik deviance REMLdev
  1794 1807 -893.2     1794    1786
Random effects:
  Groups   Name        Variance Std.Dev.
  Subject  (Intercept) 1378.18  37.124
  Residual              960.46  30.991
Number of obs: 180, groups: Subject, 18

Fixed effects:
             Estimate Std. Error t value
(Intercept) 251.4051     9.7459   25.80
Days         10.4673     0.8042   13.02


HPDinterval(VarCorr(fm1.mcmc,type="varcov"),prob=0.95)
         lower     upper
[1,] 307.8815  868.9589
[2,] 862.6868 1363.5824
attr(,"Probability")
[1] 0.95

Cheers,

Jarrod
On 11 Aug 2011, at 17:36, Emma Jones wrote: