Censoring

A person is censored if the event had not occurred by the end of their observation: they are still studying when data collection ends, or they dropped out of the survey earlier. All we know is that the event lies after that time or never occurs. Kaplan-Meier uses this information too; the person counts as at risk until the last observation. In the chart, censored people appear as small vertical ticks on the curve.

Example: A student still enrolled after 6 semesters is censored at time 6.

Caution: Do not leave censored people out and do not count them as events: both bias the curve. The method assumes that censored people have the same further risk as the others (non-informative censoring).