Regression paths from a second-order ... PreviousNext
Mplus Discussion > Structural Equation Modeling >
 hai hong li posted on Monday, July 23, 2007 - 2:11 pm
Dear Linda,

I have a SEM model that has regression paths from a second-order factor to other factors.

Can Mplus deal with it?

How to differentiate the regression paths from the paths from the second-order factor to its first-order factors?

Do you know any references of papers that involve this kind of complex model?

hai hong
 Linda K. Muthen posted on Monday, July 23, 2007 - 7:12 pm
You can differentiate them by using the ON option. I don't know of any references.
 jtw posted on Thursday, July 11, 2013 - 8:29 am
Hello there,

I'm attempting to estimate a second order factor model to capture domain-specific attitudes (first-order) and general attitudes (second order) and then using this model to predict an outcome. I am familiar with models that can accomplish this when regressing the outcome on the second order factor exclusively. However, what I am really interested in theoretically is whether domain-specific attitudes also contribute to the outcome above and beyond general attitudes. Thus, I would like to regress the outcome on both the second order factor as well as one or more of the first order factors. Are there any references you have for this type of model and/or do you foresee any problems with this setup?
 Bengt O. Muthen posted on Thursday, July 11, 2013 - 2:37 pm
Take a look at Gustafsson-Balke in MBR 1993 where these models are discussed in detail.
 Chris R posted on Wednesday, December 10, 2014 - 5:02 am

I just started to use MPLUS and I would like to test a model with two first-order reflective, second-order formative constructs (higher-order constructs) as independent variables.

I would like to ask how to consider the reflective and formative thinking in the model specifications in MPLUS. Currently, I am testing the model with a code like this:

V1_DIM1 BY ind1_1 ind1_2;
V1_DIM2 BY ind2_1 ind2_2 ind2_3;
V1_DIM3 BY ind3_1 ind3_2 ind3_3;

V1 by V1_DIM1 V1_DIM2 V1_DIM3;

V2_DIM1 BY ind4_1 ind4_2 ind4_3 ind4_4 ind4_5;
V2_DIM2 BY ind5_1 ind5_2 ind5_3;

V2 BY V2_DIM1 V2_DIM2;

V3 BY ind6_1 ind6_2 ind6_3;

V3 on V1 V2;

As you can see, the code does not distinguish between formative and reflective construct definitions. Could you please give me advice?

Kind regards,

 Linda K. Muthen posted on Wednesday, December 10, 2014 - 2:14 pm
See the Topic 1 course handout under formative indicators.
 Leslie Hafer posted on Thursday, March 10, 2016 - 12:31 pm
Is it wrong to regress the outcome variable on a second-order factor and the first-order factors that form the second-order factor. My outcome is an observed variable that measures retained or not retained. For example:
F1 by a b c
F2 by d e f
F3 by g h i
F4 by F1 F2 F3
Retained on F4
Retained on F1 F2 F3
I looked at the reference above for Gustafsson and Balke and that seems to only refer to the measurement model and not the structural model.
Are you familiar with any references regarding a structural model that regresses the outcome variable on both the second-order factor and the first-order factors that comprise the second-order factor? Thank you.
 Linda K. Muthen posted on Thursday, March 10, 2016 - 4:48 pm
When you do that, retained becomes a factor indicator for f4, f1, f2, and f3. The factor model is factor indicators regressed on factors.
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