Date:         Thu, 22 Jan 2009 10:08:25 -0500
Reply-To:     Gene Maguin <emaguin@buffalo.edu>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Gene Maguin <emaguin@buffalo.edu>
Subject:      Re: Curvilinear regression
Comments: To: Talk Stats <talkstats@GMAIL.COM>
In-Reply-To:  <200901220541.n0LMeDgw005916@malibu.cc.uga.edu>
Content-Type: text/plain; charset="US-ASCII"

Angela,

Ok, it sounds like you have two unconnected questions, one about mediators
and one about moderators.

I disagree with your characterization of your basic model x -> y as
'curvilinear', given that x is categorical. Your anova showed a main effect
for x and either planned comparisons or post hoc tests showed that x(1)
differed from x(2) and from x(3). That is not curvilinear. Curvilinear
implies to me that x is (quasi-) continuous variable having, for example, a
power relationship (x squared) with y. You don't have that.

Ok. Moderation is easier to do. (Let me also suggest that you take a reading
cruise through some good regression texts.) I'd suggest that you center your
moderator, V, by subtracting its mean prior to use in the regression. This
will minimize collinearity issues. Next, construct the product of x and V.
Since X is categorical with three levels, you'll need to construct two
contrast variables, xc2 and xc3, such that xc2 represents the x(2) vs x(1),
x(3) contrast and xc3 represents the x(3) vs x(1), x(2) contrast. Then
construct product variables for the xc2-V and xc3-V products. Now run your
regression.

Mediation. You definitely need to some background reading. The classic is
Baron and Kinney from the early 1980's. But, the current leaders in
mediation are David McKinnon and somebody whose name I can't think of at the
moment. Perhaps somebody else will name the person. Anyway, this is a good
google topic.

Gene Maguin


>>I am wondering who can give me the exact and very specific (fool proof)
steps on SPSS how to run regressions to test my curvilinear model and my
mediators( two continuous variable M1 and M2)
1.I ran anova before and am pretty sure there is a curvilinear relation
between my x and Y. my x is a catergorical variable with three levels , I
coded it using two dummies.
2.. I also have a moderator variable(v) continuous) supposedly to interact
with x and jointly influencing Y?

Any help on the steps to test 1. main effect of X on the curvilinear 2.
interaction between x and moderatorV? 3. steps to test mediation (two
mediators) 4. I also want graphs on interaction....

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD