Assessing statistical views of natural selection: Room for non-local causation?