Course contentsShow
← Data Science
Lesson 815 of 2,14521. Survival AnalysisFree lesson

Survival Curve Plots and Interpretation

Creating and interpreting Kaplan-Meier plots, including step functions and confidence intervals.

Survival Curve Plots and Interpretation

What you'll learn: How to create and read Kaplan-Meier plots that visualize survival probabilities over time, including their characteristic step-function shape and confidence bands.

The Visual Story of Survival

After computing the Kaplan-Meier estimator, we plot the survival function S(t) to see how the probability of "surviving" (not experiencing the event) changes over time. This creates a step function that drops at each observed event time—not smoothly, but in distinct downward steps.

Why steps? Because survival probability only changes when events actually occur. Between events, the curve stays flat. Each drop's size depends on how many individuals were at risk and how many experienced the event at that moment.

Reading the Plot

The y-axis shows survival probability (0 to 1 or 0% to 100%). The x-axis represents time. The curve starts at 1.0 (everyone "alive" at time zero) and typically decreases rightward as events accumulate.

Confidence intervals appear as shaded regions or dashed lines around the main curve, showing the uncertainty in our survival estimates. These bands usually widen over time, especially as the risk set shrinks and censored observations accumulate—we have less data to estimate survival at later time points.

Key Features to Notice

  • Steep drops: Many events happening at specific times
  • Long flat sections: Periods with no events or only censored observations
  • Wide confidence bands: High uncertainty (small risk set)
  • Tick marks or crosses: Often indicate censored observations

Think of it like watching a race where runners drop out at different checkpoints. The survival curve shows what percentage is still running at each checkpoint, dropping only when someone stops, not between checkpoints.

Key Takeaway: Kaplan-Meier plots display survival as a step function that drops at event times, with confidence intervals revealing estimation uncertainty—wider bands mean less reliable estimates at those time points.