Some difficulties and inconsistencies when using habit strength and reasoned action variables in models of metered household water conservation.