Give a "real world" (i.e. qualitative not quantitative) example of three variable X, Y and Z, for which you expect X and Y to be marginally associated but conditionally independent, controlling for Z