[R] multiple error strata in aov

[Prof Brian Ripley]

On Wed, 16 Jun 2004, Murray Logan wrote:

> I am trying to perform a model 3 ANOVA for a 2 factor (say factor A and 
> factor B) anova in which factor A is fixed and factor B is random. 
>  Therefore, the error term for the test of factor A should be the A:B 
> interaction term and the error terms for B and A:B should be the model 
> residual (within) term.  I have tried to work out how to specify such 
> error strata using aov, however, I have had little success.  Is there a 
> way to specify just the interaction (A:B) as an error term using Error() 
> within aov or provide a list of error strata.  
> Ideally as Error=list(A="A:B",B="Within", A:B="Within")

Error is documented in ?aov, which has a reference.  Have you read it?
It does contain the answer to your question.  I will not try to explain it 
here, and your attempt at specification suggests you do not understand the 
meaning of `strata' here.  The strata are function of the way the 
experiment was designed, not the AOV tables you want which should be a 
consequence of the design.

> In addition, I have tried using lme to perform this function, but again 
> without much success.

-- 
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