s-news
[Top] [All Lists]

Solution: contrasts in interaction term

To: s-news@wubios.wustl.edu
Subject: Solution: contrasts in interaction term
From: Ana Bio <a.bio@geog.uu.nl>
Date: Thu, 28 Jun 2001 14:50:05 -0700
In-reply-to: <5.0.2.1.0.20010628134033.00a7aec0@pop.geog.uu.nl>
Thanks a lot to Chuck Cleland who pointed at the function model.matrix() to see the coefficients used for a particular model.

Through this function the very easy (really obvious) answer to my question emerged.

There are no real "contrasts" in this case. The occurring factor level becomes "1" whereas all others become "0" in the equation. In practice, this means that the value for the continuous variable in the interaction term is simply multiplied by the coefficient for the occurring factor level.

Thanks again to all,

Ana

--------------------------------------------------

Original question:

Dear All
I am currently applying Generalized Linear Models for plant species-environment modelling. These models contain continuous variables, factors and an interaction term between a factor and a continuous variable. Because the variables in this interaction term only make sense as "interaction" they are modelled as
x1:x2 (as opposed to x1*x2)
with x1 = factor and x2 = continuous variable. This results in one coefficient for each factor level. My question is which contrasts are used for the factor in the interaction term. Contrasts are well documented for factors, but I cannot find a reference for a factor:continuous variable interaction.
Thanks,
Ana


<Prev in Thread] Current Thread [Next in Thread>