Curvilinear Effects PreviousNext
Mplus Discussion > Structural Equation Modeling >
 Achim Schmitt posted on Tuesday, June 02, 2009 - 7:53 am
Dear Linda,
I was wondering if there is a possibility to estimate non-linear (curvilinear) relations in SEM. It is not a growth model (as discussed previously in your Discussion Forum). I am aware of Quasi ML (Muthen, Klein), however, I am not sure if you integrated this approach in Mplus. Is there a syntax that accounts for curvilinear relations at one given point of time? Thanks for your help!
 Linda K. Muthen posted on Tuesday, June 02, 2009 - 9:13 am
Yes, latent variable interactions are available with full-information maximum likelihood using the XWITH command. See also Example 5.7 which is non-linear CFA.
 Achim Schmitt posted on Tuesday, June 02, 2009 - 3:17 pm
Linda, thanks for your prompt reply!
 Mathias Höglund posted on Wednesday, April 14, 2010 - 12:17 am
Dear Linda,

I'm wondering whether it's possible in mplus to test whether a relationship has an inverted u-shape? I'm getting an unexpected negative relationship that I would like to test and I'm trying to figure out the correct syntax for such a situation.

Always thankful for your help,
 Linda K. Muthen posted on Wednesday, April 14, 2010 - 8:01 am
The best model I can think of to test that would be a quadratic growth model.
 Mathias Höglund posted on Sunday, April 25, 2010 - 11:45 pm
Thank you Linda! I'll try that out! I haven't been working that much with LGM but I'll come to your course in Barcelona!
 Jonas Reichelt posted on Monday, September 19, 2011 - 7:58 am
Is it also possible to specify other polynomials for the relationship of two latent variables, i.e. comparing different types of nonlinear relationships to find out which specification results in the best fit?

Thank for any help,
 Linda K. Muthen posted on Tuesday, September 20, 2011 - 10:23 am
Can you be more specific about what you mean?
 Jonas Reichelt posted on Friday, September 23, 2011 - 12:40 am
I understand that the XWITH command allows to define quadratic relations.
Would it also be possible to define a relationship of 2 latents in the form of y=a0+a1*x+a2*x^2+a3*x^3 or even polynomials of higher degree? Is there any way to test this?

(I am aware that this would be theoretically justifiable in only few cases.)

 Linda K. Muthen posted on Saturday, September 24, 2011 - 9:00 am
You can do the following:

xx | x XWITH x;
xxx | x XWITH xx;
xxxx | x XWITH xxx;

Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message