Statistics Glossary → Censoring (survival analysis)
Censoring (survival analysis)concept
Not knowing exactly when the event happened for an observation; most often because the event has not yet occurred when the study ends.
Right censoring is the most common type: at the end of follow-up the subject has not yet had the event or has left the study; it is known that the event happens 'after this time' but not when. Discarding these observations or treating the time as an event time seriously biases the results.
Survival methods (Kaplan-Meier, Cox) use the information in censored observations that the event did not happen at least up to that time. The key assumption is non-informative censoring: leaving the study must not be related to the risk of the event. When the event time is only known to lie between two examinations, it is called interval censoring.
An everyday example
Patients who started at a clinic in 2020 stop being followed in 2025. For a patient who recovered in 2023 the time is known; for a patient still under treatment in 2025 it is only known to be 'at least 5 years'. This second patient is a censored observation and is not removed from the dataset.
In MerQur
The 'Event' variable and 'Event code' option in the survival analyses (event = 1, censored = 0); interval-censored survival analysis works with lower and upper bound columns.
Analyses where this concept appears
Cox Regression (Proportional Hazards Model)