Here is a bit of Don's example, and then a closing comment about what
I had said, about interpreting coefficients.
On 10 Jan 2000 00:20:59 -0800, [EMAIL PROTECTED] (Donald F.
Burrill) wrote:
< snip, much example >
> the coefficients b3 and b4 are indistinguishable from zero. Dropping
> SEX and SMOKES from the model, we then have
>
> PULSE2 = a + b1*PULSE1 + b2*RAN + b5*SEX*RAN + b6*SMOKES*RAN + error
>
> Now some folks get unhappy at thme idea of retaining an interaction
> (e.g., SEX*RAN) without one of its main effects (e.g., SEX); but look
- Yes, I am typically unhappy, but this is a fairly special equation,
anyway. It illustrates a point, rather than proving one ...
> (I'm not at all sure I understand what you want to mean by
> "strictly interpretable". Did you have some kind of "non-strictly" in
> mind? "Fuzzily", perhaps?)
By "strictly interpretable", I guess that I meant that I could read a
regression equation as the first step in figuring out what tests and
coefficients are important. I could read a printout and be *sure*
that the coefficients were telling me something useful about the
sample and the data. You get that with orthogonal coefficients, or
with coefficients that you examine in stages -- like looking at the
main effects before entering in the interactions.
Or, to draw another distinction, the *data* would be speaking to me.
I can get a message from reading the above, but it has to be a message
from a *statistician*, whose work and explanation I am willing to
accept.
Don has demonstrated that a set of coefficients can be drawn up, with
specific contrasts, to illustrate some outcome. But that is after
studying group means, and so on. I will try to keep that possibility
in mind, the next time I want to comment on coefficients.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
