Has your data had this sort of non normal distribution look (as in depression example)

Worth considering Robust regression and Bootstrapping and not just ignoring outliers 🙂

<<<NEW METHODS>>>