Formative Model Indicator Directions

[Anniekent@gmail.com]

Hello,

I am working on my dissertation, which involves measurement models with formative 1st and 2nd order measures. I’d greatly appreciate your help with the following questions:

1) Theoretically, in a formative model, arrows should point from the indicators toward the latent variable. However, when I did this, the AVEs for the 1st order indicators are not available in the report. How would I assess the quality of the model then?

2) Some discussions in earlier posts suggest making the arrows pointing toward the indicators (as in a reflective scale) rather than latent variables. Is it a valid method? Is there any reference?

3) I adopted the method of adding factor scores of the 1st order LVs to the 2nd order LVs to run the model. Shall the arrows be pointing toward the LV as well?

4) Is there any reference on developing formative measurement models with PLS? I am really confused about how to assess the overall quality of the model, criteria on which items to keep in the model and what to report.


Thank you very much for your kind help!

[Diogenes]

Hi Annie,
1) In formative model we don´t expect that the indicators have correlation one each other, they don´t have a common cause (or common factor or "latente" variable that explain their correlation). For this reason we don´t use AVE, communality or relyability analysis as we do in reflective model.
To assess formative models take a look at:
Index construction with formative indicators: An alternative to scale develo… Adamantios Diamantopoulos; Heidi M WinklhoferJMR, Journal of Marketing Research; May 2001; 38, 2.
ROSSITER, John R. The C-OAR-SE procedure for scale development in Marketing. International Journal of Research in Marketing. v.19, p.305-335, 2002.
special issue JBR --> viewtopic.php?t=657

2) It is not correct. If the LV is formative the arrows goes from MV to LV.

3) There are 4 kinds of combination (1st and 2nd order X reflective and formative). See JARVIS, C.B.; MACKENZIE, S. B.; PODSAKOFF, P. M. A critical review of construct indicators and measurement model misspecification in Marketing and Consumer Research. Journal of Consumer Research, v. 30, n. 2, p.199-218, set. 2003.
--> and some discussion (in Portuguese yet) in http://mee-pls.googlegroups.com/web/Var ... %20PLS.doc

4) In formative LV we should have a census of indicators and take care about multicolinearity, see references above for more details.

Best regards
Bido

[Anniekent@gmail.com]

Dear Professor Diogenes Bido,

Thank you very much for your reply and the references recommended. Gudergan, Ringle, Wende and Will (2008) has great insights on how to determine whether a scale should be formative vs. reflective. However, what if a scale is known to be a formative one. Is there any criterion to assess the quality of the model? How about the communality index and the redundancy index? Can these indicies tell the quality of the measurement model?

Thank you very much again! I really appreciate your insights.

[Diogenes]

Hi Annie,
1) statistically not, we should assess the multicolinearity for interpretation matters and to avoid (or to explain) nonsignificant weights, where was expected that were significant.
2) Communality index and the redundancy index are used in reflective mode only.
Best regards.
Bido