|
|
 |
| Mediation Analysis with 2 Outcomes |
 
|
| Message/Author |
|
|
|
|
Hi,
I want to conduct a mediation analysis which has the following:
X-leading to Mediator (Y) and the Mediator leading to two outcomes (Y1 and Y2). Y1 is a continuous variable and Y2 is a dichotomous variable. The Mplus Version is 8.0.
I am aware about how to conduct the above mediation analysis separately, i.e.,
1) X---M--Y, where Y is continuous
2) X---M--Y, where Y is dichotomous
However, is there a path analysis procedure/ program through which I can run employ both the outcome variables together as a single model. |
|
|
|
|
| Mplus doesn't give the counterfactually-defined effects for the binary Y is you have 2 Y's. But you can analyze them one at a time - the only disadvantage is if you have missing data on one while not on the other. |
|
|
|
|
Dear Prof. Muthen,
Thanks a lot for your response and the advise.
I will follow them. |
|
|
|
|
Dear Prof. Muthen,
I have a query about another model.
I want to run a moderated mediation analysis where
a) the moderator influences the relationship between the independent and mediator
b) the outcome variable is a count while all the other variables are continuous.
c) The data is panel and specifically firm-level financial data.
I want to know whether this can be done using MPLUS Version 8.0. Can you suggest any reading/s which will help me understanding the above. I know moderated mediation using MPLUS and Hayes procedure.
However, panel data and outcome being count. is this possible and what model or command should I use?
Best Wishes,
Saurabh |
|
|
|
|
With a mediation model with a count outcome, the effects should be evaluated using the counterfactual approach mentioned in reference on our Mediation page:
http://www.statmodel.com/Mediation.shtml
See especially the 2016 book by Muthen, Muthen, Asparouhov and also the 2011 paper by Muthen.
I don't know how many time points you have, but with panel data and several time points, other modeling issues arise. |
|
|
| Back to top |
|
|
