The approach is not to test a particular theory but rather to draw on a large number of models to produce a list of variables of interest in an exploratory fashion (see Atkinson and Brandolini, 2006, for a discussion of the clear limitations with such an approach).