Formative model evaluation

Joensuu · Post by **Joensuu** » Wed Oct 26, 2011 8:12 am

I'm working with a model with 7 formative measured variables (forming inner model). Each of them have three or four manifest variables measured by 1-7 likert scale. How can I evaluate the model? I used the bootstrap sampling method and got t-values which all are over 2,5. There are no problems. Now I'm wondering if there are any other ways to evaluate the model and fit with formative variables?

Thanks for advance

ctumelero · Post by **ctumelero** » Sun Nov 13, 2011 2:20 pm

Hello Joensuu,

I'm experiencing the same. What I can share with you for now, is that according to what I found in the literature the formative model evaluation should be done through:

1. EVALUATION OF THE MEASUREMENT MODEL (INDICATORS)
1.1 Multicollinearity (VIF) - through SPSS Regression. You can regress all VL indicators on the basis of one indicator of the dependent variable of your model. VIF loads below 10 should be expected.
1.2 Significance of the weights (Student's t test) - OK, you've done through bootstrap.

2. EVALUATION OF THE STRUCTURAL MODEL (CONSTRUCTS)
2.1 R2 of endogenous latent variables
2.2 Estimation of the factor weights
(2.1 and 2.2 are both from PLS Algorithm. It´s the same evaluation required for models with reflective indicators)

I hope I have helped and I also hope that someone else can add something.

Regards,

zia · Post by **zia** » Sun Dec 04, 2011 2:08 pm

I M FACING THE PROBLEM HOW TO WRITE THE FORMATIVE MODEL; I HAVE INDICATORS REFLECTIVE BUT THEY ARE REFLECITON OF SUB LATENT VARIABLES WHICH ARE 4 IN NUMBER AND .... ALL FOUR ARE THEN FORMING (FORMATIVE FOR ONE MAJOR LATENT VARIABLE... HOW TO WRITE IT PLEASE

Joensuu · Post by **Joensuu** » Tue Dec 13, 2011 12:07 pm

If I understood rigth you can't invert reflective indicators to formative mode. If that is in case just click the right button of your mouse and select "invert measurement model". Does it help?