Binary and continuous outcomes PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Joy Choi posted on Tuesday, April 10, 2018 - 8:16 pm
Hi Drs. Muthen,

So I am testing the mediating effects of M on the relationship between X and Y.

X = two continuous variables
M = one continuous variable
Y consists of 3 latent variables, 1 continuous variable and 1 categorical (yes/no, I will name it as "Vol") variable.

And I hope to use FIML option for my missing data.

1) Do I HAVE TO include the syntax "CATEGORICAL = Vol;"? When I do this, it says " THIS MODEL CAN BE DONE ONLY WITH MONTECARLO INTEGRATION."

2) So I tried Estimator = WLSMV; then it says "The model is not supported by DELTA parameterization. Use THETA parameterization." Thus I added the syntax, parameterization = theta; it worked, but the model fit became so much worse than when I did not say that Vol is categorical....

Could you please tell me whether #2 is the only way to go if there is no option for the "Vol" variable?

Thank you!
 Bengt O. Muthen posted on Wednesday, April 11, 2018 - 4:35 pm
1) Yes. Using Montecarlo integration is fine. Note that ML(R) does not give overall model fit measures with categorical.

2) Fine.

3) Yes - or use Estimator = Bayes where you also get an overall fit measure.
 Joy Choi posted on Thursday, April 12, 2018 - 8:42 pm
Thanks. I used the Bayesian estimation and it worked, but I am supposed to report the model fit (e.g., χ2 (xx) = x, p = .05, CFI = .xx, TLI = .xx, RMSEA = .xx, and SRMR = .xx.) which is not presented by the estimation. Are there any other options? I am sorry that I am familiar with this estimation. Thanks again!
 Joy Choi posted on Thursday, April 12, 2018 - 9:10 pm
I forgot to add this:

(3 continuous DV and 1 binary DV)

Can I use Estimator = MLR; INTEGRATION = MONTECARLO; along with MODEL CONSTRAINT: options instead of using Estimator = Bayes to test the mediating effects?

In this case, do I not need to report the model fit (e.g., CFI, TLI, RMSEA, and SRMR)?
 Bengt O. Muthen posted on Friday, April 13, 2018 - 4:22 pm
Our Bayes output doesn't yet have CFI etc - instead, use the Bayesian PPP. See the Bayes papers on our website, such as

Muthén, B. (2010). Bayesian analysis in Mplus: A brief introduction. Technical Report. Version 3. Click here to view Mplus inputs, data, and outputs used in this paper.
download paper contact author show abstract

WLSMV is the only estimator that give CFI etc. in this case.
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: