I am hoping someone can give me some advice and references about testing a multiple mediation model using PLS. I have a model with two IVs (representing a 2x2 experimental design), three mediator latent vars, and a latent DV.
I would like to test hypotheses that the effects of the IV's on the DV are being mediated (partially or fully) by the three intervening/mediator vars. This requires computing the indirect effects of the IV's on the DV and the associated significance. However, the procedure to do this is not clear to me.
I experimented with calculating the indirect effects and associated significance using a path model that included direct paths from the IV's to the DV as well as the hypothesized mediation pathways. This enabled me to get both the total effects and direct effects from a single PLS estimation. Then I boostrapped 1000 samples and used the bootstrapped data to compute the indirect effects for each sample. I then calculated the Std Deviation of the 1000 samples and calculated the significance of the indirect effects using the original indirect effect.
I am worried that this computational procedure is not correct. In particular, I am not certain what model specification(s) should be used to calculate the total effects and direct effects.
Any advice will be greatly appreciated. Also, any references for papers that discuss the procedure for testing multiple mediator models using PLS would be great. There are several papers I have found that deal with testing single mediator models, but none focusing on multiple mediators.
Thanks!