Negative indicator loadings and weights

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Tommy
PLS Junior User
Posts: 1
Joined: Fri Apr 15, 2022 1:16 pm
Real name and title: Tommy K

Negative indicator loadings and weights

Post by Tommy »

Hi there,
first of I hope I'm posting this in the correct section.

I have a model with 3 reflective and 2 formative constructs.
1)
Initially, I was confused about the outer loadings for my reflective indicators. All of the reflective indicators (except 2) had a negative outer loading. I made sure that this is not a reverse coding issue. I then proceeded with deleting a formative indicator with a VIF over 3. This then "fixed" all outer loadings for my reflective indicators (they are all positive now). Could someone explain this to me? Why does a formative indicator impact the reflective indicators of other constructs? "The reliability of one construct is independent of and calculated separately from that of other constructs" (Urbach & Ahlemann, 2010, p.18). Am I mixing something up here?

2)
Even after deleting 2 formative indicators with a VIF of 3.1 and 2.5, the loadings and weights of formative indicators have both negative and positive values. How could this be explained? When checking for the significance of weights and loadings of formative indicators none of them are significant. I believe that a general multicollinearity issue may be the case for my formative indicators. Nevertheless, what would be the best way to proceed here to achieve meaningful results?

Thanks a lot and kind regards
Tommy
Post Reply