Date: Thu, 30 Apr 1998 09:32:19 -0500
Reply-To: "Nichols, David" <nichols@SPSS.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Nichols, David" <nichols@SPSS.COM>
Subject: Re: monte carlo simulations & continuous variable interactions
An interesting question. What comes to mind would be to break the problem
into pieces. That is, I don't think the method that allows you to create
variables with given covariance/correlation structures is going to be able
to accommodate the other issues unless you can translate them into that
framework. Thus, you need to be able to specify the correlational structure
desired. You could approach this from the perspective of partial
correlations and work backwards to solve for the necessary bivariate
correlations (the solutions would probably not be unique, so you'd have to
make choices among alternative sets of values). I'm not positive that you'd
be able to accommodate the desire to have x3 be a product of x1 and x2
without further trickery though. Perhaps this is a start.
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
nichols@spss.com
----------
From: Joshua Sacco [SMTP:josh@PILOT.MSU.EDU]
Sent: Monday, April 27, 1998 11:30 PM
To: SPSSX-L@UGA.CC.UGA.EDU
Subject: monte carlo simulations & continuous variable interactions
hi,
i'm wondering if anyone has come across any SPSS matrix code (or has any
ideas how on a good approach for this) to generate
correlation/covariance matrices consisting of 4 continuous variables x1,
x2, x3 & y where x3 is the product of x1 & x2 and x3 predicts y after
controlling for x1 & x2. that is, i want to generate correlation
matrices representing a situation in which two continuous variables
interact to predict a third continuous variable. I've seen the code to
generate simulations using regular (i.e., those with non-interacting
variables) correlation matrices, and Aguinis (1997) references some code
to do this for the subgroup (i.e., dichotomous moderator) case, but I
don't have any good ideas about how to do this for continuous variables.
I'd also like to be able to manipulate the extent of the moderation
effect if possible.
tia,
Josh
| 
