>I have the folliwing question in mind:
>Suppose we want to do a regression or an ANOVA but we look histograms and
>residual/fitted values and the non-normality is obvious. In this case,
>someone can suggest to transform the initial values in using a Box-Cox
>transformation for exemple, and hope to obtain normality and so hope
>to do good tests and inference.
.........................
Normality is of some importance in the precise levels of the tests, but
as I have previously stated, this does not mean what those applying it
think it means. But a more important item in using these models is
linearity. This is really the key assumption, and is guaranteed to be
changed if one uses any non-linear transformation whatever.
--
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
Phone: (317)494-6054
hrubin@stat.purdue.edu (Internet, bitnet)
{purdue,pur-ee}!snap.stat!hrubin(UUCP)