1. CB-SEM and PLS-SEM
1. But CB-SEM assumes normality of data distributions, which is seldom met in social sciences
research.
2. In contrast,PLS-SEMisnon-parametricand not only works well with non-normal distributions,
3. There are alternative approaches to CB-SEM (other than the maximum likelihood algorithm)
that are designed to work with non-normal data, but they require large sample sizes. In our
example,the CB-SEMmethodresultedina substantial loss of indicator variables in an effort to
achieve acceptable GOF. At the same time, the PLS-SEM method enabled retention of many
more indicator items, which supports both measurement and structural theory development.
If the theorybeinginvestigatediswell established,andthe measurementiseffectivelyexecuted,then
CB-SEM often works well.