Skip to content
Prev 3653 / 20628 Next

Another case of -1.0 correlation of random effects

Kevin E. Thorpe wrote:
Regarding lme4a: how do I obtain it?  I guess that comes down to, what 
is the repository to give to install.packages()?  Does it require a 
different Matrix package than the one I have, which is, 0.999375-33 and 
if so, how do I not break my current lme4/Matrix combination?

By the way, the problems with these data get stranger.  In a different 
outcome from the same trial the following results from a model fitting 
attempt.

lmer(iAUC~Treatment+Dose+(Treatment|Subject)+(Dose|Subject),data=insulin)
Linear mixed model fit by REML
Formula: iAUC ~ Treatment + Dose + (Treatment | Subject) + (Dose | Subject)
    Data: insulin
   AIC  BIC logLik deviance REMLdev
  1956 1982   -968     1983    1936
Random effects:
  Groups   Name         Variance   Std.Dev.   Corr
  Subject  (Intercept)  4.2678e-02 2.0659e-01
           TreatmentOat 1.6115e+07 4.0144e+03 0.000
  Subject  (Intercept)  3.0430e+08 1.7444e+04
           Dose         1.5173e+06 1.2318e+03 -1.000
  Residual              3.1907e+07 5.6486e+03
Number of obs: 96, groups: Subject, 12

Fixed effects:
              Estimate Std. Error t value
(Intercept)   40142.4     5146.7   7.800
TreatmentOat   1340.3     1634.7   0.820
Dose          -2675.1      405.5  -6.597

Correlation of Fixed Effects:
             (Intr) TrtmnO
TreatmentOt -0.079
Dose        -0.922  0.000

As you can see, I get a 0 correlation within one set of random effects 
and -1.0 in the other.  Also, the fact the fixed effects estimates are 
huge makes me suspicious.

Note that if I drop the treatment portions and fit the Dose model to 
only one treatment, the correlation is again -1.0 and the fixed effects 
are similar.

Thread (28 messages)

Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 9 Ben Bolker Another case of -1.0 correlation of random effects Apr 9 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 9 Ken Knoblauch Another case of -1.0 correlation of random effects Apr 9 Ben Bolker Another case of -1.0 correlation of random effects Apr 9 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 9 Viechtbauer Wolfgang (STAT) Another case of -1.0 correlation of random effects Apr 9 Jarrod Hadfield Another case of -1.0 correlation of random effects Apr 9 Ken Knoblauch Another case of -1.0 correlation of random effects Apr 9 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 9 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 12 Douglas Bates Another case of -1.0 correlation of random effects Apr 12 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 12 Viechtbauer Wolfgang (STAT) Another case of -1.0 correlation of random effects Apr 12 Douglas Bates Another case of -1.0 correlation of random effects Apr 12 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 13 Ben Bolker Another case of -1.0 correlation of random effects Apr 13 Douglas Bates Another case of -1.0 correlation of random effects Apr 13 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 13 Ben Bolker Another case of -1.0 correlation of random effects Apr 13 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 13 Douglas Bates Another case of -1.0 correlation of random effects Apr 13 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 13 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 15 Andrew Dolman Another case of -1.0 correlation of random effects Apr 16 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 16 Andrew Dolman Another case of -1.0 correlation of random effects Apr 16 Kevin E. Thorpe Another case of -1.0 correlation of random effects Apr 16