I have build a model with reflective and formative constructs. I know how I can test the reliability and validity of the reflective constructs, but I have some issues with testing the reliability and validity of the formative constructs.
I follow the recommendations of Petter, Straub and Rai (2007), which are based on an extensive literature review:
Assess construct validity
- Use principal component analyse to examine the item weightings for measures
Can I also conduct a bootstrap procedure in smartPLS to examine the significance? Petter et al. give 2 options: 1 eliminate or keep the items. But why should you eliminate the items, because it impacts the content validity.
Evaluate reliability
- Examine multicollinearity to determine if VIF <3.3.
I have no issues with a high VIF, so my reliability is fine?
Analyse formative constructs via component-based SEM (e.g., PLS)
- Specify formative constructs in the research model using mode C.
o What is mode C?
- Examine model weights for formative measures.
o Based on what criteria can I examine the model weights?
- Assess R2 values for endogenous variables in the structural model.
o Which R2 are valid?
Title: Specifying Formative Constructs in Information Systems Research
Authors: Petter, Straub, Rai
Journal: MIS Quarterly
Year: 2007