|Date: ||Mon, 17 Jan 2005 13:32:08 -0500|
|Reply-To: ||Andrew Walsh <firstname.lastname@example.org>|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>|
|From: ||Andrew Walsh <email@example.com>|
|Subject: ||multiple regression: interaction|
|Content-Type: ||text/plain; charset=ISO-8859-1; format=flowed|
i would like to pose a question regarding interactions in multiple
regression for which i would appreciate some advice.
i have a hypothesized model where: (1) physical symptom distress, (2)
perceived social support, and (3) anxiety interact to influence levels
of depression(DV) (a 3-way interaction). so my question is: is it
appropriate to jump right to the 3-way interaction without examining the
2-way interactions? i ask this for 2 reasons. first, when i enter the
three 2-way interactions into my model along with the 3-way interaction,
two of the 2-way interactions are significant (perceived social support
x anxiety is not significant) and the 3-way interaction is not
significant. however, the 3-way interaction is significant when it is
in the model without the three 2-way interactions.
thanks for any and all advice!