Survey measurement error in population case/control studies: retrospective reports of risk factor exposures