Dear PLS-experts,
I hope you can help me on the problem I have.
I have a model with 2 latent variables (Portal loyalty and Overall loyalty), each determined by 2 independent variables. The latent variables can be connected. When I connect these latent variables, item A - part of portal loyalty - only just surpasses the 0,70 item-loading cut-off, and therefore should be retained.
If I delete or reverse the relationship between Portal loyalty and Overall loyalty, item X no longer surpasses this cut-off, and should be deleted before relationships can be analysed.
What is the reason behind this? Does this have anything to do with how factor-loadings are calculated in PLS? Are only modeled relationships estimated, and does this affect item-loading outcomes?
Thanks in advance,
William