Changing Signs of Factor Loadings

[Tobias Gress]

Hi all:

I came across an effect that I don’t know how to interpret. Depending on where I put paths between LVs in my model, my factor loadings change. I tested all the constructs, using CFA and established reliability and validity before I started modelling in SmartPLS. Also the factor loadings of all the items in my main model are high and significant.

When I change paths between LVs (i.e. model an interaction effect) the factor loadings change. In some cases 2 of 4 previously positive loadings become negative and significant.

Can anyone tell me how I must interpret this effect when analysing the structural model?

Kind regards
Tobias