Mediated Moderation with SPSS output

[Kmaxine]

I am running some final analysis on mediation and moderation tests. I have
both under different scenarios and was now looking to see if I have med-mod
or mod-med. I have all the relavant articles, but wondering how that
translates to reading the unstand betas from the SPSS output. For example,
with mediation (partial) the significance drops. What do I look for with
med-mod or mod-med.
Any suggestions?
Kmax

[statisticsdoc]

Kmax,

If you are applying the Baron-Kenney criteria for mediation, then you will
need to utilize the Sobel test for the significance of change in beta
weights when a mediator is introduced into the model.  SPSS does not have
perform the Sobel test directly, nor does any software that I know of.
However, it is possible to copy the relevant beta weights into an Excel
spreadsheet to perform the test.

HTH,

Steve

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


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Kmaxine
Sent: Saturday, June 09, 2007 2:19 PM
To: [hidden email]
Subject: Mediated Moderation with SPSS output


I am running some final analysis on mediation and moderation tests. I have
both under different scenarios and was now looking to see if I have med-mod
or mod-med. I have all the relavant articles, but wondering how that
translates to reading the unstand betas from the SPSS output. For example,
with mediation (partial) the significance drops. What do I look for with
med-mod or mod-med.
Any suggestions?
Kmax

[F. Gabarrot]

Hello,

According to Muller et al. (2005), you cannot utilize the Sobel test with mod-med or med-mod because you need to enter both the mediator and the interaction between the moderator and the mediator as predictors. As the Sobel test is a 1 df test, it cannot be used with mod-med/med-mod.

You will find the analytic procedure to use for, and what to look for, in the Muller et al. (2005) paper.

Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852-863.

The analytic models for mediated moderation and moderated mediation are the following :

Step 1) Y = β1.0 + β1.1*X + β1.2*Mo + β1.3*X*Mo + ei
Step 2) Me = β2.0 + β2.1*X + β2.2*Mo + β2.3*X*Mo + ei
Step 3) Y = β3.0 + β3.1*X + β3.2*Mo + β3.3*X*Mo + β3.4*Me + β3.5*Mo*Me + ei

And, Prototypical Mod-med is characterized as follows: β1.3, β2.3 and β3.4 (case 1) or β1.3, β2.1 and β3.5 (case 2) must be significant, and |β1.3| > |β3.3|

Prototypical Med-mod, however, will be characterized as follows : β1.1 must be significant -but not β1.3- β2.3 and β3.4 significant, or β2.1 and β3.5 significant. |β1.3| > |β3.3|.

Hope this helps.

(I hope this is understandable, english is not my mother tongue).

statisticsdoc wrote

Kmax,

If you are applying the Baron-Kenney criteria for mediation, then you will
need to utilize the Sobel test for the significance of change in beta
weights when a mediator is introduced into the model.  SPSS does not have
perform the Sobel test directly, nor does any software that I know of.
However, it is possible to copy the relevant beta weights into an Excel
spreadsheet to perform the test.

HTH,

Steve

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


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[e-mail]]On Behalf Of
Kmaxine
Sent: Saturday, June 09, 2007 2:19 PM
To: [e-mail]
Subject: Mediated Moderation with SPSS output


I am running some final analysis on mediation and moderation tests. I have
both under different scenarios and was now looking to see if I have med-mod
or mod-med. I have all the relavant articles, but wondering how that
translates to reading the unstand betas from the SPSS output. For example,
with mediation (partial) the significance drops. What do I look for with
med-mod or mod-med.
Any suggestions?
Kmax

[statisticsdoc]

Kmax,

The following link to the Muller, Judd & Yzerbyt (2005) may be helpful.

http://209.85.165.104/search?q=cache:4W0-2GYsj_kJ:www.research.plymouth.ac.u
k/healthpsych/wp-content/uploads/2006/02/muller-05-mediatingmoderation.pdf+m
ediated+moderation&hl=en&ct=clnk&cd=2&gl=us

My previous posting may have introduced some confusion into the discussion.
Dr. Gabarrot is correct in stating that the Sobel test is not the critical
test of med-mod or mod-med.  I gather that you have used the Sobel test, so
you will recall that it is useful in detecting whether the relationship
between the initial distal predictor and the dependent variable is
significantly attentuated by the mediator, even if there is still a
significant relationship between the distal predictor and the dependent
variable (i.e., partial mediation).

In any case, as is the case with regular mediation, med-mod and mod-med
require the synthesis of findings from multiple runs.  No direct procedure
exists in syntax for testing a med-mod or mod-med model.

In your data, is Structural Equation Modeling a feasible alternative?  For
example, AMOS can be used to test whether the mediational path is invariant
across groups.

HTH,

Steve


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


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Kmaxine
Sent: Saturday, June 09, 2007 2:19 PM
To: [hidden email]
Subject: Mediated Moderation with SPSS output


I am running some final analysis on mediation and moderation tests. I have
both under different scenarios and was now looking to see if I have med-mod
or mod-med. I have all the relavant articles, but wondering how that
translates to reading the unstand betas from the SPSS output. For example,
with mediation (partial) the significance drops. What do I look for with
med-mod or mod-med.
Any suggestions?
Kmax

[Paul Ginns-2]

Hi all,

 

You can get SPSS macros for the Sobel test (including bootstrapping options) at

 

http://www.comm.ohio-state.edu/ahayes/ under SPSS and SAS macros (the indirect.sps is Andrew's recommended one for the standard mediation model and modmed.sps can do moderated mediation).

 

Cheers,

 

Paul

 

 

Date:    Sun, 10 Jun 2007 01:13:18 -0400

From:    Statisticsdoc <[hidden email]>

Subject: Re: Mediated Moderation with SPSS output

 

Kmax,

 

If you are applying the Baron-Kenney criteria for mediation, then you will need to utilize the Sobel test for the significance of change in beta weights when a mediator is introduced into the model.  SPSS does not have perform the Sobel test directly, nor does any software that I know of.

However, it is possible to copy the relevant beta weights into an Excel spreadsheet to perform the test.

 

HTH,

 

Steve

 

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

 

 

-----Original Message-----

From: SPSSX(r) Discussion [mailto:[hidden email]]On <mailto:[hidden email]%5dOn>  Behalf Of Kmaxine

Sent: Saturday, June 09, 2007 2:19 PM

To: [hidden email]

Subject: Mediated Moderation with SPSS output

 

 

I am running some final analysis on mediation and moderation tests. I have both under different scenarios and was now looking to see if I have med-mod or mod-med. I have all the relavant articles, but wondering how that translates to reading the unstand betas from the SPSS output. For example, with mediation (partial) the significance drops. What do I look for with med-mod or mod-med.

Any suggestions?

Kmax

 

------------------------------

 

Date:    Sun, 10 Jun 2007 01:50:31 -0700

From:    "F. Gabarrot" <[hidden email]>

Subject: Re: Mediated Moderation with SPSS output

 

Hello,

 

According to Muller et al. (2005), you cannot utilize the Sobel test with mod-med or med-mod because you need to enter both the mediator and the interaction between the moderator and the mediator as predictors. As the Sobel test is a 1 df test, it cannot be used with mod-med/med-mod.

 

You will find the analytic procedure to use for, and what to look for, in the Muller et al. (2005) paper.

 

Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852-863.

 

The analytic models for mediated moderation and moderated mediation are the following :

 

Step 1) Y = β1.0 + β1.1*X + β1.2*Mo + β1.3*X*Mo + ei Step 2) Me = β2.0 + β2.1*X + β2.2*Mo + β2.3*X*Mo + ei Step 3) Y = β3.0 + β3.1*X + β3.2*Mo + β3.3*X*Mo + β3.4*Me + β3.5*Mo*Me + ei

 

And, Prototypical Mod-med is characterized as follows: β1.3, β2.3 and β3.4 (case 1) or β1.3, β2.1 and β3.5 (case 2) must be significant, and |β1.3| >

|β3.3|

 

Prototypical Med-mod, however, will be characterized as follows : β1.1 must be significant -but not β1.3- β2.3 and β3.4 significant, or β2.1 and β3.5 significant. |β1.3| > |β3.3|.

 

Hope this helps.

 

(I hope this is understandable, english is not my mother tongue).

 

 

 

[George_L]

Hello,

I am trying to find my way through the whole med-mod and mod-med methodology, and I stumbled upon your post.

Would that mean, because the Sobel test is a 1df test there would not be any df left to test the mod-med and med-mod (as they take into account more than one relationship?).

Sincerly yours,
George

F. Gabarrot wrote

Hello,

According to Muller et al. (2005), you cannot utilize the Sobel test with mod-med or med-mod because you need to enter both the mediator and the interaction between the moderator and the mediator as predictors. As the Sobel test is a 1 df test, it cannot be used with mod-med/med-mod.

You will find the analytic procedure to use for, and what to look for, in the Muller et al. (2005) paper.

Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852-863.

The analytic models for mediated moderation and moderated mediation are the following :

Step 1) Y = β1.0 + β1.1*X + β1.2*Mo + β1.3*X*Mo + ei
Step 2) Me = β2.0 + β2.1*X + β2.2*Mo + β2.3*X*Mo + ei
Step 3) Y = β3.0 + β3.1*X + β3.2*Mo + β3.3*X*Mo + β3.4*Me + β3.5*Mo*Me + ei

And, Prototypical Mod-med is characterized as follows: β1.3, β2.3 and β3.4 (case 1) or β1.3, β2.1 and β3.5 (case 2) must be significant, and |β1.3| > |β3.3|

Prototypical Med-mod, however, will be characterized as follows : β1.1 must be significant -but not β1.3- β2.3 and β3.4 significant, or β2.1 and β3.5 significant. |β1.3| > |β3.3|.

Hope this helps.

(I hope this is understandable, english is not my mother tongue).

statisticsdoc wrote

Kmax,

If you are applying the Baron-Kenney criteria for mediation, then you will
need to utilize the Sobel test for the significance of change in beta
weights when a mediator is introduced into the model.  SPSS does not have
perform the Sobel test directly, nor does any software that I know of.
However, it is possible to copy the relevant beta weights into an Excel
spreadsheet to perform the test.

HTH,

Steve

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


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[e-mail]]On Behalf Of
Kmaxine
Sent: Saturday, June 09, 2007 2:19 PM
To: [e-mail]
Subject: Mediated Moderation with SPSS output


I am running some final analysis on mediation and moderation tests. I have
both under different scenarios and was now looking to see if I have med-mod
or mod-med. I have all the relavant articles, but wondering how that
translates to reading the unstand betas from the SPSS output. For example,
with mediation (partial) the significance drops. What do I look for with
med-mod or mod-med.
Any suggestions?
Kmax