Matteo thanks! I've noticed many of causal models use OLS. But OLS explains just linear dependence. I assume most of the dependencies are non-linear. Is it common in practice to use non-linear models insted of OLS (boostings for example)?