Increasingly, people argue that the Hayes method may not be accurate for testing mediation. Some recommend to only use SEM methods for testing mediation, see for example the special feature topic on ao mediation in Organizational Research Methods (http://journals.sagepub.com/toc/orma/20/4)

Thanks Robert. But how to treat this model withe multilevel and multiple (serial) mediation (TFL-->PS-->TL-->IWB)?

I'm not sure about serial mediation in PROCESS (check all the models), but in SEM-based methods you can model this.

For the team level antecedents, I'm assuming you have the same team for multiple observations? In that case, your IV and mediators will not be independently distributed and you will potentially have correlated errors. That is a whole different problem from the modelling of the mediation itself.

Wilfred,

It appears that you collected the data on TFL, TL , PS and IWB ll at the individual level.

Would maintaining the unit of analysis at the individual level not avoid the multi-level challenges ? Is there theoretical justification for an individual level of analysis as opposed to a multi-level (team -> individual) model ?]

Silburn