Detecting survival effects of socioeconomic status: Problems in the use of aggregate measures