[R-sig-ME] lmer vs lmer2 vs AD Model Builder RE

[dave fournier]

As a further thought on robust estimation for
mixed models I fit  the full data set with
a model based on a 95% mixture of a normal distribution
and a 5% Cauchy distribution  I obtained the
parameter estimates:

   real_b          -1.9405e+000 9.5807e-002
   real_b           1.6089e+000 3.2403e-002
   real_b           2.0168e-002 1.3129e-003
   real_b           1.8537e-001 1.1414e-001
   real_b          -1.7127e-002 7.7816e-003

compared to the estimates from R with the
outlier removed.

(Intercept) -1.948119   0.095877  -20.32
   Height       1.640650   0.032800   50.02
   Age          0.019379   0.001310   14.79
   InitHeight   0.143977   0.111043    1.30
   InitAge     -0.014618   0.007501   -1.95

They are almost identical for statistical purposes.
Also by restricting the parameterization of the full covariance matrix 
for the random effects slightly we are able to estimate it
which speaks to the stability of  the AD Model builder Random Effects
software.




David A. Fournier
P.O. Box 2040,
Sidney, B.C. V8l 3S3
Canada
Phone/FAX 250-655-3364
http://otter-rsch.com