Hey all,
I really need your help!
In my model I receive several path coefficients far greater than 1. I 've heard that this is possible (article from Jöreskog) but I would like to understand why, and what I can do to avoiod it.
I would be really happ if you could help me out with this!
Thanks!
Simone
Path coefficients greater than 1
Path coefficients greater than 1
Simone Fucks
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hey Simone,
firstly you should check if you work with standardized values.
In the next step you should check if there are any outliers in your dataset. Outliers could often produce strange results.
Often there is a problem with multicolinearity which produce higher path coefficients than 1. In this case it could be very useful to work with another weighting scheme. In my experience the factor weighting scheme working much better in cases of multicolinearity.
Best regards,
Christian
firstly you should check if you work with standardized values.
In the next step you should check if there are any outliers in your dataset. Outliers could often produce strange results.
Often there is a problem with multicolinearity which produce higher path coefficients than 1. In this case it could be very useful to work with another weighting scheme. In my experience the factor weighting scheme working much better in cases of multicolinearity.
Best regards,
Christian
Dear Christian,
thanks for your help!
I thought that PLS does the standardization automatically when running the algorithm?
Multicolinearity testing is a good point, I will try it right away. But there is one further problem: I included a moderator term (as a product term of two other constructs), and of course, the multicollinearity is high in that case. any suggestions?
Thanks
Simone
thanks for your help!
I thought that PLS does the standardization automatically when running the algorithm?
Multicolinearity testing is a good point, I will try it right away. But there is one further problem: I included a moderator term (as a product term of two other constructs), and of course, the multicollinearity is high in that case. any suggestions?
Thanks
Simone
Simone Fucks