Dear all,
Is it possible in smarpls, to determine whether a relationship goes from A to B, or B to A?
Most often I naturally see the same correlation when reversing the arrow between to LV. However, is it possible, without considering the logical and chronological order, to mathematical determine in which direction the relationship is strongest?
Br
Determining direction of a relationship
-
Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi,
In SmartPLS, a model with just two LV will have the same path coefficient for: (A <-- B) or (A --> B)
I think that the “correct, best or usual” way is justifying the direction based on theory, but I understand your point. Maybe you could find some help in d-sep or d-separation of Pearl and Shipley:
Pearl, J. (2000a). Causality: models, reasoning and inference. Cambridge: Cambridge University Press.
Pearl, J. (2000b). The Logic of Counterfactuals in Causal Inference. Forecast, 95(450), 1–8.
Pearl, J. (2007). The Mathematics of Causal Inference in Statistics. Biometrics, 19–26.
Shipley, B. (2004). Cause and Correlation in Biology: A User’s Guide to Path Analysis, Structural Equations and Causal Inference. Cambridge: Cambridge University Press.
Best regards,
Bido
In SmartPLS, a model with just two LV will have the same path coefficient for: (A <-- B) or (A --> B)
I think that the “correct, best or usual” way is justifying the direction based on theory, but I understand your point. Maybe you could find some help in d-sep or d-separation of Pearl and Shipley:
Pearl, J. (2000a). Causality: models, reasoning and inference. Cambridge: Cambridge University Press.
Pearl, J. (2000b). The Logic of Counterfactuals in Causal Inference. Forecast, 95(450), 1–8.
Pearl, J. (2007). The Mathematics of Causal Inference in Statistics. Biometrics, 19–26.
Shipley, B. (2004). Cause and Correlation in Biology: A User’s Guide to Path Analysis, Structural Equations and Causal Inference. Cambridge: Cambridge University Press.
Best regards,
Bido
Hi Bido,
Thank you for your reply.
All the models I work with will have more than two LV. usually between 5-12. However, sometimes it can be difficult to determine the relationship.
E.g consider these three variable I (image), P(product), R(Re purchase)
My model would then be (I-->R) (P-->R)
However, if I were to test the relationship between I and P.
The image of a brand, could theoretical, effect my satisfaction with the product, and my willingness to re-purchase the product. However, my satisfaction/experience with a product, could also effect my perception/image of the brand of that product. This is why I was curious about how to determine in which direction the relationship is strongest.
I will try and look into your suggested articles.
Thank you.
BR
Thank you for your reply.
All the models I work with will have more than two LV. usually between 5-12. However, sometimes it can be difficult to determine the relationship.
E.g consider these three variable I (image), P(product), R(Re purchase)
My model would then be (I-->R) (P-->R)
However, if I were to test the relationship between I and P.
The image of a brand, could theoretical, effect my satisfaction with the product, and my willingness to re-purchase the product. However, my satisfaction/experience with a product, could also effect my perception/image of the brand of that product. This is why I was curious about how to determine in which direction the relationship is strongest.
I will try and look into your suggested articles.
Thank you.
BR