Have you experienced this?
When you try a structure equation analysis with Lisrel, Amos or SmartPLS, some requirements are simply not met. The distribution assumptions are not met? Non-linear relations might be there, or worse: moderating
variables could be interfering, but you don’t know which ones?
Which variables are actually affecting which other variables? And in which direction? Question after question that might not be answered by your literature review on proven theories... Shouldn’t ‘real’ research be different?
The goal is that you only use the knowledge and theories that you can
truly back up. The analysis method should do the rest. These methods
should also be flexible enough to show you where, and which, non-linearities and moderation effects which you were not previously aware of are at play.
