Introduction to the Kaplan-Meier Estimator
What you'll learn: How the Kaplan-Meier method estimates survival probabilities even when some observations are censored.
What Is the Kaplan-Meier Estimator?
The Kaplan-Meier (KM) estimator is a non-parametric method that calculates the probability of surviving past certain time points using censored data. Remember from earlier lessons that censoring occurs when we don't observe the actual event for some subjects—they drop out, the study ends, or they're lost to follow-up.
Why Do We Need It?
Imagine you're tracking customer churn over 12 months. Some customers cancel (the event happens), but others are still active when your study ends (right-censored). You can't just ignore the censored customers—they provide valuable information about survival up to their last known time point. The KM estimator cleverly incorporates both groups.
The Core Idea
The KM estimator works by:
- Ordering all event times from earliest to latest
- Calculating survival probability at each event time by multiplying the probability of surviving that specific moment by the cumulative survival up to that point
- Accounting for censored observations by including them in the "at-risk" count up until they're censored, then removing them
Think of it like tracking runners in a marathon where some drop out (events) and others leave the course but might finish later (censored). The KM estimator uses all available information to estimate what fraction survives at each milestone.
The result is a step function that drops at each event time, creating the characteristic "survival curve" you'll visualize.
Key Takeaway: The Kaplan-Meier estimator provides survival probabilities over time by intelligently using both complete and censored observations, making it the gold standard for basic survival analysis.