Date: Tue, 20 Mar 2001 14:54:17 -0300
Reply-To: "Hector E. Maletta" <hmaletta@OVERNET.COM.AR>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Hector E. Maletta" <hmaletta@OVERNET.COM.AR>
Subject: Re: plotting interactions
Content-Type: text/plain; charset=us-ascii
Sonya,
one possible way is to plot the regression of your dependent variable
against the predicted value (from an equation containing only your
independent variables TR and TM), split by your moderating variable SM.
This would give you only one chart, with two sets of points (one color
for low SM, another for high SM). You can show the regression lines for
both sets, and get the points hidden with a suitable choice of marker or
color.
The steps for this are as follows:
1. Create SMBINARY with values 1 and 2 for high and low SM.
2. Choose SPLIT FILE BY SMBINARY.
3. Do a REGRESSION of SU with TR and TM as indep.variables. Choose to
save the predicted values as a new variable, say SUPRED.
4. Cancel the SPLIT FILE setting.
5. Do a scatter diagram of SU by SUPRED, split by SMBINARY.
6. Edit the graph. In the editor, choose to show the regression lines
and edit the color/markers to hid the points, so that only the lines are
visible. Alternatively you can keep the points visible, but this may
result in a confusing graph if the points are many (remember they are
from two sets, represented by two colors/markers).
The resulting regression lines will show the effect of the moderating
variable SM. The line with low SM, I suppose, will have a steeper slope,
if SM plays a moderating influence.
Hope this helps.
Hector Maletta
Universidad del Salvador
Buenos Aires, Argentina
Sonya Premeaux wrote:
>
> This is a multi-part message in MIME format.
>
> ------=_NextPart_000_0029_01C0B130.4A3E0920
> Content-Type: text/plain;
> charset="iso-8859-1"
> Content-Transfer-Encoding: quoted-printable
>
> I need to plot interaction effects for a regression. Can anyone help? =
> My data looks something like this:
>
> SM SU TR TM SE
> 1 1 1 1 1
> 2 2 2 2 2 =20
>
> My dep variable is SU. SM is a moderator of the relationships between =
> SU and the other variables. I want to split SM into High and Low and =
> plot regression equations. I would end up with 2 regression equations =
> on each chart so that my dep variable is on my Y axis, one of the IVs is =
> on the X axis and there is a regression line for high SM and low SM. =
> Can this be done? If so, how? If not, any suggestions on how to do it =
> elsewhere? Thanks.
>
> ------=_NextPart_000_0029_01C0B130.4A3E0920
> Content-Type: text/html;
> charset="iso-8859-1"
> Content-Transfer-Encoding: quoted-printable
>
> <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
> <HTML><HEAD>
> <META content=3D"text/html; charset=3Diso-8859-1" =
> http-equiv=3DContent-Type>
> <META content=3D"MSHTML 5.00.2919.6307" name=3DGENERATOR>
> <STYLE></STYLE>
> </HEAD>
> <BODY bgColor=3D#ffffff>
> <DIV><FONT size=3D2>I need to plot interaction effects for a =
> regression. Can=20
> anyone help? My data looks something like this:</FONT></DIV>
> <DIV> </DIV>
> <DIV><FONT size=3D2>SM SU =
> TR =20
> TM SE</FONT></DIV>
> <DIV><FONT size=3D2>1 =20
> 1 =
> 1 =20
> 1 1</FONT></DIV>
> <DIV><FONT size=3D2>2 =20
> 2 =20
> 2 2  =
> ; =20
> 2 </FONT></DIV>
> <DIV> </DIV>
> <DIV> </DIV>
> <DIV><FONT size=3D2>My dep variable is SU. SM is a moderator of =
> the=20
> relationships between SU and the other variables. I want to =
> split SM=20
> into High and Low and plot regression equations. I would end up =
> with 2=20
> regression equations on each chart so that my dep variable is on my Y =
> axis, one=20
> of the IVs is on the X axis and there is a regression line for high SM =
> and low=20
> SM. Can this be done? If so, how? If not, any =
> suggestions on=20
> how to do it elsewhere? Thanks.</FONT></DIV></BODY></HTML>
>
> ------=_NextPart_000_0029_01C0B130.4A3E0920--
| 
