Kathy -
An example of where an intercept maybe should be excluded is this....
Consider modeling the time it takes to board an airplane by the number of
passengers. If there are 0 passengers, we would expect it to take 0 minutes
to board the plane. Here we may want to exclude the intercept term.
In short, unless you have strong theoretical reason to do so, it is
generally advised to leave the intercept term in the model.
Peter Link
VA San Diego Healthcare System
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]]On Behalf Of
Kathy McKnight
Sent: Wednesday, June 06, 2007 10:46 AM
To:
[hidden email]
Subject: excluding the intercept from a GLM
Hi all,
I'm going to be demonstrating several models for analyzing change data, and
one model I'm using is a pre/posttest ANCOVA. In that model, I use 8th grade
math scores as a covariate, along with gender & taking advanced math in 8th
grade (Yes/No variable) to predict a math gain score (12th grade math score
minus 8th grade math score). I realize the problems with such a model, but
for now, I want to skip those and go directly to my question about excluding
the intercept from the model. In the help menu, SPSS indicates that if I can
assume that the data pass through the origin, I can exclude the intercept in
the GLM I'm testing. My question is this: does that mean I would assume my
data pass through 0 (meaning a score of 0 on the math test, I assume)? I'm
guessing that this means that somewhere back in the annals of time, math
achievement scores would pass through 0, which doesn't seem to me to be a
necessary assumption for this model. I would really appreciate an example of
when one would EXCLUDE the intercept from a model. I appreciate any and all
help :)
Thanks much,
Katherine McKnight
Locked
