Factor loadings and SEM PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Rina posted on Saturday, July 07, 2007 - 5:04 pm
Hello Dr. Muthen and Muthen,

I am doing a model like this:

F1 by a1 a2 a3;
F2 by b1 b2 b3;
F3 by c1 c2 c3 c4;

F2 on F3;
F2 on F1;

They are all continuous variables.
The results showed O.K. model fit. But the factor loadings of F3 look counterintuitive.
The STDXY loadings for c1 c2 c3 and c4 are .82, .90, .20, and .23, respectively.
I am wondering why there is such big difference between the factor loadings of c1 c2 and those of c3 c4? Could there be any problems? Would it affect the results of the entire model?
I did CFA for F3, the loadings also look like that. Could you tell me what I can do to improve my model fit, or, how to ensure that my model is identified? Thank you!
 Linda K. Muthen posted on Sunday, July 08, 2007 - 8:23 am
This would indicate that the reliabilities of two items are lower than for the other two. It is not a reflection on the fit or identification of the model.
 Rina posted on Wednesday, July 11, 2007 - 9:10 am
Hello Linda,

A follow up of your response, what did you mean by "the reliabilities of two items"? C1 C2 C3 and C4 are observed variables each contains 5 to 14 items. The internal reliability (Cronbach Alpha) are:
C1: .59;
C2: .76;
C3: .60;
C4: .48.
I don't see that Alphas of C3 is lower than C1. Could you clarify it? Are we talking about the same "reliability"?

Another questions, if this is a issue caused by the data itself, is there anyway to deal with it? Thank you!
 Linda K. Muthen posted on Wednesday, July 11, 2007 - 10:24 am
I am referring to reliability in the factor analysis sense -- variance explained for each indicator by the factor.

This issue is due to the data and it is dealt with by statistical modeling. Less reliable items are given less weight.
 Bee Jay posted on Sunday, March 25, 2012 - 2:57 pm
I am also interested in looking at StdYX for path coefficients. I can get a value by using STANDARDIZED, but I want the table with the p-values, like on page 643 of the manual. When I type in that command though "STANDARDIZED (STDYX);" I get an error saying STDYX is an unrecognized option for output. What is the correct option? Somewhere online it had said just "STDYX" but that didn't work for me either.

Thanks!
 Linda K. Muthen posted on Monday, March 26, 2012 - 8:04 am
You must be using an older version of the program where the option is STANDARDIZED. You cannot use specific options.
 Bee Jay posted on Monday, March 26, 2012 - 5:56 pm
Thank you for your response. That's unfortunate for me. So there's no way to obtain p-values?

Also, I have a problem with some negative path coefficients (StdYX values). In this scenario, is the best option to constrain the path to be greater than 0 using MODEL CONSTRAINT?
 Linda K. Muthen posted on Monday, March 26, 2012 - 6:03 pm
If you don't come automatically, then you can't get them without looking them up in a z-table.

Path coefficients can be positive or negative. They are regression coefficients.
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: