[讨论][推荐] Linear and Curvilinear Relationships

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Hi all.

I am attempting to test a mediation hypothesis. There is a significant
correlation between my mediator variable and my DV. However, when I further
examined the nature of this relationship by regressing the DV onto the
mediator and including a quadratic term, my output indicates that both the
linear model and the quadratic model are significant, and that actually my
quadratic model accounts for slightly more variance than the linear model
(however, both Rs are small). What is the importance of this finding? Does
this mean it is inappropriate to test a mediation hypothesis given the
existence of a non-linear relationship between the mediator and DV? Or can I
go ahead with the mediation and just note that as a limitation??

Taylor Poling
Industrial/Organizational Psychology Program
University of Tennessee, Knoxville
408 Stokely Management Center
Knoxville, TN 37996-0545
(865) 974-1676

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：relationship Curvilinear relations relation Linear 推荐 讨论 Linear Curvilinear

Taylor,

When you speak of the quadratic model, I assume that this model includes
both the mediator and the squared value of the mediator.

You can still test the mediation hypothesis using the Baron-Kenny criteria.
You should see how much the direct relationship between the distal variable
(the one that is being mediated) and the outcome variable is diminished when
include both the mediator and squared value of the mediator. That is, to
what extent is the unique variance of the distal variable with the outcome
variable diminished when the mediator and squard value of the mediator are
added to the regression. You can still apply the Sobel test to assess how
much the direct effect of the mediator is diminished.

You should also examine the relationship between the distal variable and
both the linear and quadratic elements of the mediator.


HTH,

KS


For personalized and professional consultation in statistics and research
design, visit
www.statisticsdoc.com