Interpreting unweighted effects coding interactions.

Hi everyone,

I think I have a simple question, but I would appreciate it if anyone
can provide an explicit answer (oh, and I've already checked the Cohen
and Cohen 2003 book, while they have answers *near* my question, they
don't answer it or address it specifically).

Okay, so say I'm doing a simple OLS Regression, and I have two
dichotomous variables and I want to test as well for an interaction
between them. Further, I have made both of these variables unweighted
effects codes (i.e., -1, 1, instead of the 'traditionally' 0, 1) as I
want to generalize in terms of the "mean of the means" of the two
groups.

Presuming I DO find a statistically sig. interaction term, how would I
properly interpret this in plain English? Further, how would I
properly set up the interaction terms for two unweighted effects? If
they were simply dummies I would just multiply them per normal, but
with unweighted effects that would (I think) create an odd outcome....
(ex: -1 x -1 = 1, 1 x 1 = 1 <-- that just doesn't seem right).

As an example... say my DV is "Liking your boyfriend score", and my
two IVs are (1) Gave flowers/ No flowers and (2) Gave chocolates/ No
chocolates... and in this case, I found my interaction term to be
positive and significant (GF likes you even more when you give both,
beyond what you'd expect by simply adding the terms).

I understand how to interpret interaction terms, simple dummys as well
as continuous... but I'm getting my head wrapped up trying to make
sure I understand *precisely* what I am seeing when I interpret the
interaction of two unweighted effects codes. I'm hoping to have a
precise understanding

Hope my question was clear, and I thank anyone who offers insights.
If anything is ambiguous or sloppy in my explanation, I'll try to
clarify.

.

Relevant Pages
Loading