Dear all,
In my structural model, there are two constructs that have reciprocal relationships with each other. That is, when I change the causal direction it is still significant. For example, A predicts B significantly and when I change the causal direction from B to A, it is still significant. Does it statistically make sense?
Regards
Seyed
Do reciprocal relationships statistically make sense?
-
yaserbanihashem
- PLS Junior User
- Posts: 3
- Joined: Tue Sep 17, 2013 4:55 am
- Real name and title:
Do reciprocal relationships statistically make sense?
Seyed Yaser Banihashemi
PhD Candidate
School of Civil Engineering
The University of Sydney
PhD Candidate
School of Civil Engineering
The University of Sydney
-
suprapto
- PLS Junior User
- Posts: 4
- Joined: Mon Oct 21, 2013 9:57 pm
- Real name and title:
- Location: Netherlands
Dear Seyed,
I have a more or less similar situation as yours. It would depend on the theory. If a strong theory does exist to support A --> B or B --> A then I think you can base on that theory and justify either direction. Or simply use the results to test/compare which theory/concept is stronger in your empirical setting.
The problem becomes complicated when there is no unifying strong theory or when we simply want to explore those possible reciprocity (which is also make sense in practice we found two aspects/constructs are intertwined to each other). Since no reciprocal relation is allowed, there is a possibility to model this reciprocal relation as a second order reflective construct (as a global higher order construct that exists to represent all inter-correlated constructs). By modeling this, we operationalize that there is a C as a global factor for both A and B where A <--> B.
Mohammad
I have a more or less similar situation as yours. It would depend on the theory. If a strong theory does exist to support A --> B or B --> A then I think you can base on that theory and justify either direction. Or simply use the results to test/compare which theory/concept is stronger in your empirical setting.
The problem becomes complicated when there is no unifying strong theory or when we simply want to explore those possible reciprocity (which is also make sense in practice we found two aspects/constructs are intertwined to each other). Since no reciprocal relation is allowed, there is a possibility to model this reciprocal relation as a second order reflective construct (as a global higher order construct that exists to represent all inter-correlated constructs). By modeling this, we operationalize that there is a C as a global factor for both A and B where A <--> B.
Mohammad
Reflect, think, support and act
