Hi all,
I am currently working on mediation analysis and effects decomposition. Since one of my main variable is a latent variable and I have different indirect paths, I just opted for structural-equation model. However, just for a robustness check, I would like to compute the mediation also with a standard Ols approach and then thinking to use the Sobel test for checking the significance.

Therefore, the idea is the following:
Code:
reg log_income X P // where X is a matrix of individual characteristics (mostly personality traits) and P is father's social class 

reg log_income X P E // where E is the mediating variable i.e. education level of the individual
The first regression represents the total effect and subtracting the second direct effect I can obtain the indirect effects. However, it seems it's not available a Sobel test in Stata anymore. How can I test the significance of the indirect effect in this case? Or is there any other way for implementing this "robustness check"?
Thanks for your help