On econometric inference and multiple use of the same data