Non-informative censoring is an assumption underlying which method?

• A. Kaplan-Meier estimator
• B. Logistic regression
• C. Poisson regression
• D. Linear regression

Answer: (A)

Non-informative censoring means that the reason and timing of censorship are independent of a person’s risk of the event, given what has been observed. The Kaplan-Meier estimator relies on this by treating censored individuals as still at risk up to their censoring time and then excluding them from risk thereafter. This allows the estimator to compute the survival probability at each observed event time without bias from why someone stopped being followed. If censoring were informative—for example, if sicker individuals were more likely to be censored—the estimated survival curve could be biased because the censored group would not have the same risk profile as those who remain under observation. The other methods listed do not inherently handle time-to-event censoring in the same way; logistic, Poisson, and linear regression model different outcomes (binary, counts, or continuous) and do not provide a survival function under censoring without additional survival-specific techniques.