[R] nested aov: plot available?

[Prof Brian Ripley]

On Thu, 11 Dec 2003 Pascal.Niklaus at unibas.ch wrote:

> Hi all,
> 
> I wonder whether, for an anova with multiple error strata, it is possible to
> produce the same diagnostoc plots than with a single-stratum anova.
> 
> I can extract the residuals for each stratum with e.g.
>   > resid(split1.aov[["block:plot"]])
>   > resid(split1.aov[["Within"]])
> and then produce qqnorm plots etc manually, but is it possible to get all the
> plots (residuals vs fitted, cook's distance etc) by stratum with a single plot
> command as with the single-stratum call to aov ? Or is there a reason why I
> wouldn't want to do this?

One usually would not want to look at all the strata in the same way.
Plots such as Cook's distance are in any case uninteresting for a fixed
design with no high-leverage points (one hopes), linearity is not an issue 
and so on.

_You_ can of course write your own plot method to do this.  I doubt if 
more that a handful of R users would want such a method, and quite 
possible only one would want it to do as you suggest.

-- 
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