85
Noninformative Censoring
In the presence of noninformative censoring, the
risk set in any interval should look like a random sample of the
population at risk
•Estimates
of the conditional probability of survival for the intervals
should be uncorrelated
