Computing the Kaplan-Meier Product
What you'll learn: How to manually calculate survival probabilities using the Kaplan-Meier product-limit formula step by step.
The Core Idea
The Kaplan-Meier estimator calculates survival probability at each time point by multiplying together conditional survival rates. At each event time, you compute the probability of surviving that specific moment (given you've survived until then), then multiply all these probabilities together to get cumulative survival.
The formula is: S(t) = ∏(1 - dᵢ/nᵢ) for all times i ≤ t
Where:
- dᵢ = number of events (deaths/failures) at time i
- nᵢ = number at risk just before time i (those who haven't yet experienced the event or been censored)
- The product symbol ∏ means "multiply together"
Step-by-Step Calculation
Example: You're tracking 5 customers. Events occur at days 3, 5, and 7.
-
Day 3: 1 event occurs, 5 customers at risk
- Conditional survival = (5-1)/5 = 0.80
- S(3) = 0.80
-
Day 5: 1 event occurs, 4 still at risk
- Conditional survival = (4-1)/4 = 0.75
- S(5) = 0.80 × 0.75 = 0.60
-
Day 7: 1 event occurs, 3 at risk
- Conditional survival = (3-1)/3 = 0.667
- S(7) = 0.60 × 0.667 = 0.40
At day 7, the estimated probability that a customer survives is 40%.
Why the Product?
Think of it like a series of gates: to survive until day 7, you must pass through day 3 AND day 5 AND day 7. Since these are sequential conditional probabilities, we multiply them together.
Key Takeaway: The Kaplan-Meier estimator multiplies conditional survival rates at each event time to produce cumulative survival probabilities, updating the "at-risk" count as events and censoring occur.