Time-to-event data record how long until something happens: death, relapse, recovery, dropping out.
Why it matters
Ordinary averages fail because some people have not had the event by the end of the study.
Key ideas
- Two columns: time and event status (1 = event, 0 = censored)
- The time starts at a clear origin (diagnosis, enrolment)
- Follow-up lengths differ between people
- Methods: Kaplan–Meier, log-rank, Cox
- Never drop censored people
Real-life research example
60 cancer patients followed for up to 5 years. 25 died; 35 were alive at last contact.
When to use
- Outcome is "time until event"
When NOT to use
- Don't average follow-up times of those who had the event only
- Don't turn it into yes/no without considering time
Mini example
Patient 1: 12 months, event = 1 Patient 2: 60 months, event = 0 (alive at end)
Quick takeaway
Time + event status = survival data.
Try it yourself
Enter times 5, 8, 12, 12, 20, 25 with events 1, 1, 0, 1, 0, 1 in the Kaplan–Meier tool. How many events and censored are there?
Model answer
There are 4 events and 2 censored observations. Censored people still contribute time until they were last seen.