This study examined the interrelations between cognition and mortality in a nationally representative sample of the U.S. population aged 70 and older. A distinct feature of this dataset is the gradual passing away of the participants over seven waves of longitudinal data collection (1993-2006). To account for this particular type of attrition in the modeling of longitudinal cognitive data, three different statistical models were used: latent curve models, change point models, and joint growth and survival models. After these, survival times for these participants were predicted using traditional survival analysis and exploratory data mining including survival trees and random forests. The results showed that episodic memory declined considerably with age, while crystallized intelligence was largely stable until about age 77 and even the decline after this point was relatively slow, which was in agreement with the aging theory of intellectual abilities (Horn & Cattell, 1966). There did not seem to be a death-driven process that was related to the deterioration in these two abilities. Consistent with many earlier findings, cognitive levels were found to be negatively associated with mortality for both episodic memory and crystallized intelligence, but only the memory slope was found to be associated with mortality. Results from different methods consistently demonstrated that cognition was a useful predictor of mortality above and beyond many known risk factors such as demographics, comorbidities, risk behaviors and functional status. The prognostic value of cognition can be potentially useful in informing clinical and policy decision making.