Latent Variable based on correlation ... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Louie P. Cagasan Jr. posted on Friday, December 02, 2011 - 3:03 am
Hi,

I am interested in examining five variables,and based on theory, they are not supposed to correlate with each other. However, common results show that correlation existed between the five variables. I hypothesize that another variable is the reason for the existence of these correlations.

Is it possible to estimate a latent variable based on the correlation results using Mplus?
 Linda K. Muthen posted on Friday, December 02, 2011 - 6:26 am
I'm not clear on your question. You can model correlated factors in Mplus. But I don't know what "estimate a latent variable based on the correlation results" means.
 Louie P. Cagasan Jr. posted on Friday, December 02, 2011 - 3:59 pm
Hi, thanks for responding. I'll try to rephrase my question. Is it possible to do the following using Mplus?

r12 -> F1
r13 -> F1
r23 -> F1

Is possible to connect the curve lines (modeled correlations) to a factor? It is as if the modeled correlations define the factor. I am also not sure if my question is conceptually sound or possible; but I hope you can enlighten me on this.

Thanks!
 Bengt O. Muthen posted on Saturday, December 03, 2011 - 8:14 am
It seems like you are thinking in terms of the r variables being "formative" indicators of the factor f1. This model needs an outcome - a variable that is influenced by f1. Otherwise the model is not identified. Much has been written about formative - see for instance articles by Bollen. We discuss how to do it in Mplus a bit at the end of our Topic 1 short course, see the handout on our web site.
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: