Negative path coefficients and multicollinearity

[lebu]

Hello,

this is my first time to use smartPLS and structural equation modelling in general. I already solved many "beginner problems" by searching this forum, but there are still some questions left:

1) Some significant weights are negative in my PLS-model although the few negative items among the indicators were correctly recoded and there should not be a negative relation to the construct. I also checked a version without the negative items to look if they are the reason, but nothing changed.
Are there other reasons to explain negative paths? I do not know how to change it, but do not have an explanation either.

2) As far as I know formative indicators should be checked for multicollinearity e.g. with VIF and "Konditionsindex". As I have a third order model (job satisfaction) with formative dimensions (e.g. working conditions) which are based on first order factors (e.g. equipment, burden...) with reflective indicators I am not sure which indicators to use in the linear regressions to check multicollinearity. As the formative dimension has one item (overall dimension satisfaction) I do not know if only this counts or also the reflective indicators on the first level? So how to check multicollinearity for the formative dimensions with one indicator?

I hope that was not too confusing that I discouraged you to answer!
Any hint would be great especially to question 1.

Leonie

[viswadatta]

A negative path coefficient refers to a negative influence of the influencing construct on the dependant construct. See if you can explain this using any theoretical backup.