Robert Bell points me to this blog by Austin Frakt explaining problems in interpreting regressions that control for intermediate outcomes. As Robert notes, we discuss these issues in chapters 9 and 10. But Frakt’s example is a good one.
Be careful what you control for . . . you just might get it!
Great post title! Hard not to agree with Frakt’s point that having Medicaid is, in general, far better than being uninsured.