Running WILD: the case for exploring mixed parameter sets in sensitivity analysis