SEM to 3 mode data
Message/Author
 Kosugi Koji posted on Sunday, January 22, 2006 - 4:12 pm
I am conducting an exploratory positioning analysis of Semantic Differential Data.
The analysis was developed by Toyoda and Saito. The details appear in an address as below.

www.littera.waseda.ac.jp/faculty/tyosem/saito/imps2004.ppt

And Macro for SAS is shown ( the number of the factors is fixed to 2)

http://www.littera.waseda.ac.jp/faculty/tyosem/jpa67/oyaku/posmac.txt

It is like doing secondary factor analysis, a primary factor expresses a characteristic of a target, and a second factor expresses individual differences.

This analysis needs the following constraints.

To primary factor
-factor patterns between targets are equal
-Intercepts between targets are equal
-sum of factor averages of primary factor is zero
-sum of factor variance of primary factor is 1
Next,about the passes from a secondary factor to primary factor
-Absolute values of a pass coefficient do not exceed 1

Because I do not have SAS, I want to construct it in Mplus.
But I have no idea how to write the constraints, especially last three.

Thank you.
 bmuthen posted on Sunday, January 22, 2006 - 7:36 pm
Model Constraint can be used for the last three, except the last one which involves an inequality and will need the forthcoming Version 4.0. See the Version 3 User's Guide for Model Constraint. Here, the sum of functions of parameters being constrained to zero, can be written as one function (one factor average) being equal to the negative of the sum of the other functions.
 Kosugi Koji posted on Monday, January 23, 2006 - 12:18 am

It was possible to constraint a pass coefficient and a intercepts equally.

However, I do not understand how I should have expressed a factor average and variance.
In MODEL CONSTRAINT, it have to use a label, but how should set a label for average and variance of a factor in MODEL?
 Linda K. Muthen posted on Monday, January 23, 2006 - 8:53 am
Please send the formulas for factor average and variance that you are trying to replicate along with your input, data, output, and license number to support@statmodel.com.