PLS-POS and Mode A/B

[ahg]

Hello everyone,

Brief background and simple question:

We have performed PLS-POS and have three groups that can be theoretically justified. We cannot perform any kind of MGA because we cannot establish measurement invariance, but we want to explain the three different groups separately (completely different paths to purchase intention). We have 3 formative constructs measured with Mode B. The problem is that we have several multicollinearity issues with different indicators across groups, resulting in negative weights in some cases. We cannot delete any indicator because the problematic ones differ across groups (something that might be expected from the nature of the three groups).

My question is: can we proceed to perform the analysis directly with the same groups just switching the formative constructs to Mode A (the results are good) or are we supposed to re-run PLS-POS again with the formative indicators in Mode A, and proceed with the analysis again with the new resulting groups? My guess is the second choice, because of how the POS groups were formed from a Mode B model, but I would like a confirmation.

Best,

Ángel.

[cringle]

Sounds good; just double click on the formative latent variable and, then, you can select Mode A. Re-running the PLS-POS analysis sounds to some extent plausible. But do you have a good explanatory variable that matches your PLS-POS groups very roughly? In that case, I would use the explanatory variable directly with your Mode A constructs.

Best regards
CR

[ahg]

That makes sense. Thanks, Christian!