I found some good explanations of what the Box Cox transform for anyone who happens upon this thread while searching for boxcox details:
This is the best explanation I found of what how the Box Cox transformation is calculated. It explains why QR decomposition is used, has nice code examples that are probably more illustrative than anything actually implemented in MASS, which I expect would have a lot of extra mechanics to make it robust. http://www.r-bloggers.com/on-box-cox-transform-in-regression-models/ http://freakonometrics.blog.free.fr/index.php?post/2012/11/13/On-Box-Cox-transform-in-regression-models As usual, Julian Faraway provides some great basic notes on using tranformations, including the Box Cox transformation: http://cran.r-project.org/doc/contrib/Faraway-PRA.pdf A useful, compact, and complete explanation of how the log likelihood function is derived: http://data.princeton.edu/wws509/notes/c2s10.html An useful discussion on the practical limitations of the Box Cox transformation, and some alternative approaches, such as mixed models. http://robjhyndman.com/hyndsight/transformations/ A much more complete and helpful help for NextMethod can be found in 5.5 of R's Language documentation. (Although I'm still not sure how to see the code in boxcox!) http://cran.r-project.org/doc/manuals/R-lang.pdf Also, a more simple and candid summary of S3 and S4 methods is here http://cran.r-project.org/doc/contrib/R_language.pdf http://www.r-project.org/conferences/useR-2004/Keynotes/Leisch.pdf [[alternative HTML version deleted]] ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
