. We then imputed cognitive status to the full HRS sample (10,903 persons, 31,936 personyears) on the basis of measures of cognitive and functional status available for all HRS respondents, thereby identifying persons in the larger sample with a high probability of dementia