Do a search for "center polynomials" in the JMP help system.
Probably you will find it easier to figure out what is going on if you
turn this option OFF in the Fit Model dialog
... although the p-value for some tests will be different (for a good
reason)
Bottom line: you have to learn exactly how the model is parameterized to
get the zen of what's going on.
Michael Bailey wrote:
I am confused about the meaning of the intercept term in jmp 7 in a
particular context (more below), because its value doesn't match up
with what I think it should.
The context is this: A dependent variable, Y, and (at least) two
predictors, X1 and X2, and their interaction. If I regress Y onto X1,
X2, and X1-cross-X2, the regression coefficients for X1, X2, and the
interaction all make sense (and the main effects are calculated by jmp
using deviation scores for X1 and X2). However, the intercept does not
match what I get when I do this regression the long way (namely, by
creating deviation scores for X1 and X2 and taking their product to
implement the interaction). Nor does it match the intercept when the
raw variables (as opposed to the deviation scores) are analyzed.
Can anyone tell me what is going on and what the Y-intercept in this
context means?
Michael Bailey
jm-bailey@northwestern.edu
|