Survival Data Mining

Main Focus

The two biggest pitfalls in SA is the considerable variation in the risk across the time interval which demonstrates the need for shorter time intervals and censoring. Censored observations occur when there is a loss of observation. This most often arises when subjects withdraw or are lost from follow-up before the completion of the study. The effect of censoring often renders a bias within studies based upon incomplete data or partial information on survival or failure times.

There are four basic approaches for the analysis of censored data: complete data analysis; the imputation approach; analysis with dichotomized data; and the likelihood-based approach (Leung et al., 1997). The most effective approach to censoring problems is to use methods of estimation that adjust for whether or not an individual observation is censored. These “likelihood-based approaches” include the Kaplan-Meier estimator and the Cox-regression, both popular methodologies. The Kaplan-Meier estimator allows for the estimation of survival over time even for populations that include subjects who enter at different times or drop out.