Ahmad Abedalfattah You can use multiple correlated independent variables in path or regression analysis to examine mediated effects. In cases in which collinearity is strong (highly correlated IVs), you may see effects of suppression or redundancy. Such effects may result in reversed signs of regression coefficients relative to the signs of the corresponding zero-order correlations, standardized regression or path coefficients > |1|, or high standard errors of regression/path coefficients). For highly redundant IVs, perhaps one of them can be dropped or the variables can be combined into a single variable (composite or summary score) if that is theoretically appropriate and meaningful.

Ahmad Abedalfattah, I agree with the colleague. You could test if you don't have a factor behind your data. You can conduct a factor analysis or a reliability analysis to check if the items can be appropriately combined and conduct the analysis using it. If the items/variables are highly correlated probably you could combine them.

Alternatively, you may try to conduct "latent profile analysis" for the highly correlated independent variables and identify "a set of profiles" (one profile may show high scores for all independent variables.

Amalia Raquel Pérez Nebra Christian Geiser

Thanks, could you please provide me reference to read

Ahmad Abedalfattah See, for example,

https://www.isonderhouden.nl/doc/pdf/arnoldbakker/articles/articles_arnold_bakker_64.pdf

Article MacKinnon DP, Krull JL, Lockwood CM. Equivalence of the medi...