RE: [R] Enduring LME confusion… or Psychologists and Mixed-Effects

[Prof Brian Ripley]

On Thu, 12 Aug 2004, Doran, Harold wrote:

>  lme fits models using restricted maximum likelihood by default. So, I
> believe this is why you have a different DF. If you include method="ML"
> in the modeling function the DF should be similar to aov.

It is REML and not ML that generalizes the classical multistratum AoV.
In any case, REML and ML use the same subspaces, just different methods 
for estimating variances within them.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595