[R] boxcox in MASS library

[Prof Brian Ripley]

On Mon, 1 Mar 2004 demiurg at post.tau.ac.il wrote:

> Help page for boxcox function in MASS library says that 
> the transformation is y^lambda, which is different from
> the Y' = log(Y) if lambda = 0 , 
> Y' = ((Y ^ lambda) - 1)/lambda otherwise I'm used to.
> 
> Is this just a help page typo ?

No.  It makes no difference to the profile log-likelihood which is what
boxcox() computes.

It looks to me as if you failed to consult the reference.  MASS is support 
software for a book, and the book has the complete details.  (This is in 
the posting guide.)

-- 
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