Moderated mediation with latent varia... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Brandon Velez posted on Wednesday, February 10, 2016 - 7:16 pm
Hello,

I am attempting a moderated mediation analysis with latent variables. I tried the approach described in UG 5.13 but I received an error message.

Thus, I used the residual centering approach described by Steinmetz, Davidov, and Schmidt (2011) to create indicators for the latent interaction variable.

1) Am I correctly using MODEL CONSTRAINTS to test the relation of disc with ih at low and high levels of coll?

2) How can I compute standardized estimates of the simple slopes (the output only provided unstandardized estimates)?

3) How can I test the conditional indirect relation of disc on dist and well through ih at low and high levels (+/- 1 SD) of coll?

ANALYSIS:
ESTIMATOR = MLR;

MODEL:
disc BY a1-a3;
ih BY c1-c3;
coll BY e1-e3;
dist BY y1-y3;
well BY y4-y6;
int1 BY x1-x9;
ih ON age;
dist well ON disc coll ih age;
disc@1;
coll@1;
ih ON disc (k1)
coll (k2)
int1 (k3);

Model Constraint:
New (lowcoll hicoll);
lowcoll = k1+k3*(-1);
hicoll = k1+k3*(+1);

MODEL INDIRECT:
dist IND disc;
well IND disc;

OUTPUT: STANDARDIZED(STDYX) CINTERVAL;

Thank you very much for any guidance you may have.
 Brandon Velez posted on Thursday, February 11, 2016 - 11:49 am
As a follow-up to my question #3 on February 10, would adding the following commands provide me with the conditional unstandardized indirect relations of disc on dist and well through ih at low and high levels of coll? Essentially, below I am assuming that multiplying the conditional direct effect of disc on ih by the effect of ih on dist or well (i.e., the conditional effect of x on a mediator X the effect of a mediator on an outcome) will yield the conditional indirect effect.

MODEL:
...
dist ON ih (k4);
well ON ih (k5);

MODEL CONSTRAINT:
New (lowcoll hicoll lowind hiindd lowindw hiindw);
lowcoll = k1+k3*(-1);
hicoll = k1+k3*(+1);
lowind = (k1+k3*(-1))*k4;
hiind = (k1+k3*(+1))*k4;
lowindw = (k1+k3*(-1))*k5;
hiindw = (k1+k3*(+1))*k5;
 Bengt O. Muthen posted on Thursday, February 11, 2016 - 6:31 pm
We request that postings not be done in two windows.

If you like you can send your failed attempt at the approach of UG ex 5.13 to Support along with your license number.

I am not familiar with the approach of Steinmetz, Davidov, and Schmidt (2011) so cannot comment on inputs related to this. Perhaps you want to contact the authors.
 Brandon Velez posted on Thursday, February 11, 2016 - 7:43 pm
Hello Dr. Bengt,

I apologize for posting in multiple windows. I will avoid doing so in the future.

I have contacted Support to determine if I could analyze the data using the approach in 5.13. I look forward to hearing back about this matter.

I feel confident that I am correctly constructing the latent variable interaction using Steinmetz et al.'s approach. However, their paper does not address testing simple slopes for significant interactions or testing conditional indirect effects.

If possible, could you please comment on whether or not I am correctly computing the simple slopes described in the MODEL CONSTRAINTS of the 2/10 post and the conditional indirect effects described in the MODEL CONSTRAINTS of the 2/11 post?
 Bengt O. Muthen posted on Friday, February 12, 2016 - 3:02 pm
Looks correct.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: