On Mon, 29 Nov 1999 19:10:16 -0600, Jim Clark <[EMAIL PROTECTED]>
wrote:
< ... >
>
> I too prefer the actual means, but some people (e.g., Rosenthal)
> have argued strongly that the residual means should in fact be
> reported to demonstrate the nature of the interaction.
Jim,
- I think that should be, "demonstrate the *direction*" -- I would
reserve "nature" for better purposes.
As Donald as pointed out about one sense of interpretation, the
"nature" of the interaction is always "disordinal" after you subtract
out the main effects -- explicitly, you have to look at the raw means
to figure out that aspect of nature, in the presence of any (strong)
main effects.
"IS it, the observed interaction, an artifact of the strong main
effect, combined with bad scaling?" If that is the case, it does not
matter what group it seemingly implies what about. And you won't get
that from the plot of residuals.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
