Dear friends, My model includes 6 constructs. 2 independent constructs, one dependent construct and 3 mediators. My hypotheses include the relationship between each construct and the mediating effects. Can I run SEM for each pair of constructs separately or single SEM for all constrcuts simultaneously to test the hypotheses?
Dear Thaneswary Raveendran
You should run the model including all the latent variables. Then, since you have mediators, analyze the direct and indirect relationships between constructs.
Bests
Thank you very much Alejandro and Lei. Really helpful as I am in the final stage of analysis.
Friends, I have one more confusion. When I run the SEM separately for each pair of constructs, the direction of relationship match with the theory. But, when I run the SEM model with all latent constructs (full model), the direction of relationship is inconsistent with the theory. For example, In the full model, the impact of transformational leadership on employee performance is negative (inconsistent with the theory and literature). Can I proceed with the full model? Can I defend this inconsistency? Please advise me.
Collinearity may change the strength (significant or non significant) or even the of the direction (+ or -) of the relation between two variables. Thus, check first if convergent and discriminant validity are guaranteed in your model.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling