I am conducting a correlational meta-analysis and want to include not only bivariate correlations but also regression coefficients as effect sizes. I know it's feasible to include multiple regression coefficients as partial or semi-partial correlations. However, I don't know how to code and analyze the coefficients from multivariate models such as SEM, path model, or latent growth model. Is that possible to include those model results? How to deal with the mediators in the path? For example, I am interested in the relation between A and B, but in the path model, there is a linear path A-C-B. I can compute the relation between A and B by multiplying the coefficients along the path but is it a correct way to code the effect size?