Mplus code for moderated mediation

[ya]

Dear list,

 

I am trying the 5 models in the following paper:

 

Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research, 42, 185-227.

 

The corresponding Mplus code can be found here:

http://www.afhayes.com/public/mplus1to5.pdf

 

I am having trouble to understand the new parameter defined in the MODEL CONSTRAINT part, for example, for model 1, the MODEL CONSTRAINT part looks like:

 

MODEL CONSTRAINT: 

      new (ind xmodval); 

      xmodval = -1; 

      ind = a1*(b1+b2*xmodval); 

 

My question is:

 

What is this xmodval? What is its function? Why a value should be give to it? The b2 is already the path coefficient of the interaction term involving the moderator, why the b2 should be multiplied with something else?

 

Thank you very much.

 

Best regards,

 

ya

[ya]

Hi Kris,

 

Thank you very much for the suggestions.

 

Have a good day.

 

Best regards,

 

---

ya

 

From: Kristopher J. Preacher

Date: 2012-11-08 22:22

To: xinxi813; SEM...@LISTSERV.UA.EDU

Subject: Re: Mplus code for moderated mediation

Ya - Yes, infinite different values (but bounded from above and below). Unless, of course, your moderator starts out being discrete, in which case it is sensible to explore only the values that are actually observed (e.g., 0 and 1).

Kris

At 01:32 PM 11/8/2012, ya wrote:

Hi Kris,
 
Thank you very much for the response.
 
If the continuous moderators are not being recoded into categorical variables, Since it is continuous, theoretically, does it mean that there will be infinite number of values  to be specified? or did you mean only the values appeared in the data set need to be specified?
 
Best regards,
 

ya
 
From: Kristopher J. Preacher
Date: 2012-11-08 09:51
To: SEMNET
Subject: Re: Mplus code for moderated mediation
Hello ya,

No continuous variables are being recoded into categorical variables. xmodval = -1 does not in general mean 1 SD below the mean of x. It is just a conditional value of x that happens to equal -1 in the example. You can use any value you want, and as many as you want, as long as they are sensible values of x. For instance, if x ranges from 2 to 15 in your sample, you could do the following, dispensing with xmodval altogether:

MODEL CONSTRAINT: NEW(ind2-ind8);
ind2=a1*(b1+b2*2);
ind3=a1*(b1+b2*3);
ind4=a1*(b1+b2*4);
ind5=a1*(b1+b2*5);
ind6=a1*(b1+b2*6);
ind7=a1*(b1+b2*7);
ind8=a1*(b1+b2*8);

Kris

At 01:38 AM 11/8/2012, ya wrote:

Hi Garett,
 
Thank you very much for the help.
 
So those -1, 0,and 1 referred to 1 SD below , equal and above the mean.
 
In this case, Does it mean the continuous moderator was manually recoded into a categorical variable, so that the calculation of the indirect effect conditioned on the moderator become possible, right? In other words, it is not possible(or not quite necessary) to calculate the indirect effect for every value of a continuous moderator, so the cut off was need. Do I understand it correctly?

 
Thank you very much.
 
Best regards,
 

ya
 

From: Garett Howardson
Date: 2012-11-07 13:15
To: SEMNET
Subject: Re: Mplus code for moderated mediation
Hi Ya, 

The modval is the value of the moderator at which the conditional indirect effects will be tested. This is just as in multiple regression when one tests the simple slopes of the interaction effects be examining the effects of the predictor on the outcome at various levels of the moderator, typically one standard deviation above and below or something lik that. In other words, the simple slopes are plotted to interpret the interaction term (if it is significant).

In moderate mediation, the same practice is followed. The indirect effects are plotted at various levels of the moderator to help interpret the meaning of the interaction term. This is what "modval" is the Preacher et al. (2007) piece - it's the conditional value of the moderator at which the indirect effects are estimated. Just like in typically regression, you can do this usine -1, 0, and 1 for the "modval" values to examine the indirect effects at values of 1 SD below the mean, the mean, and 1 SD above of the mean of the moderator, respectively. Thus, the purpose of the model constraint command and the the "modval" variable is to interpret the moderated mediation function by testing the sign and significance of the indirect effect at various levels of the moderator.

HTH,
Garett