[R] Lack of independence in anova()

[Douglas Bates]

On 7/6/05, Phillip Good <pigood at verizon.net> wrote:
> Do you or Lumley have a citation for this conclusion?   Most people go
> forward with the ANOV on the basis that the various tests are independent.
> 
> Phillip Good
> P.S.  Tests based on the method of synchronized permutations are
> independent.

Perhaps we could review the sequence of events here.  This exchange
began with your sending me a message claiming that there is a bug in
lm or anova in R because the results of your simulation were what you
expected.  I responded saying that it is unlikely that a serious bug
in such a fundamental part of R would remain undetected for such a
long time and then went further and took apart your simulation and
showed how it could be done much more effectively and also showed,
with help from Thomas Lumley and Peter Dalgaard, that your original
assumption is incorrect.

You have now made another blanket statement the "Most people go
forward with the ANOV on the basis that the various tests are
independent" and indicated that Thomas or I should provide a citation
to validate our claim.  Perhaps instead of claiming that it is
necessary for us to produce evidence in support of our claim that they
are not independent because they are based on the same denominator
mean square, you could produce a citation to back up your claim.  To
date none of your claims of the independence of the tests or the
supposed bugs in R have been substantiated.  If indeed "most people"
do as you claim then "most people" are suffering from a misconception,
a not-uncommon situation in statistics.