[R] anova.glm() and deviance/df

[Prof Brian Ripley]

On Fri, 4 Jul 2003, Henric Nilsson wrote:

> Is seems to me that R doesn't allow correct F-tests to be computed in anova.glm
> after estimating the dispersion parameter by deviance/df. Or does it?

It does when they are correct, e.g. for a Gaussian family.


-- 
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