Causal Mediation using mediate command

[Alan Acock]

In the documentation, there is a model having exercise as the treatment, bonotonin as the mediator, and well-being as the outcome. In an example of confounding variables, they say they assume randomization to the exercise vs. control group. If you had self-selection, then gender could be a confounder for exercise. Perhaps women are more likely to pick the exercise program. Stata documetation has gender as a confounding variable with a direct effect on bonotonin. Does this mean you do not need to have gender as a confounder for the Treatment? Does having gender as a confounder for the mediator and outcome mean you don't need to have it as a confounder for the treatment? I hope this is correct because I can't add it as a confounder for the treatment.