moderation effect with formative and bi-dimensinal construct

[tharwa najar]

I have a model composed of one independent variable, one dependent variable and four moderation relations. The moderators (2 bi-dimensional and 2 uni-dimensional) are reflective and the independent variable (predictor) is bi-dimensional formative construct. Based on the literature, we have to adopt the two-stage approach (Henseler & al, 2010) which stipulates;
1. The smart PLS out put generates the scores of each latent variable. Those latent variable scores are calculated and saved for further analysis.
2. The second stage is the interaction term (X ×M) which is built up as the elementwise product of the latent variable scores of X and M. After that, the interaction terms (the latent variable scores of X and M) are used as exogenous variables in a multiple linear regression on the latent variables scores of Y.
However, I have some questions here;
1. The scores of the LV are extracted before or after the test of the validity and the reliability of the main model and after eliminating the non significant items? If it’s not when will I proceed to the tests; after including the moderating constructs?
2. I have some bi-dimensional constructs. While calculating the interaction of LVS, should I use the score of the principal construct or the factors composing it. If it’s the second proposition, this will be applied for dependent and moderators constructs, both of them?
3. After calculating the interaction term of the moderators variables, how can I export it as a exogenous construct in the Smart PLS model. Should I create a new project with data base of Latent variables scores??
4. How can I interpret the significance of the moderation in this case?
5. I’m really very confused and it’s the first time that I use the smart PLS and especially the moderating effect analysis

Thank you for the response