What is the Log-Rank Test?
You'll learn: How to statistically test whether survival curves from different groups are significantly different.
The Core Idea
When you've plotted multiple survival curves (say, one for a treatment group and one for a control group), you naturally want to ask: "Are these curves really different, or could the differences just be due to chance?" The log-rank test is the most common statistical test for answering this question.
Think of it like comparing two roads through a mountain range. Both start at the same elevation, but one seems to descend faster. The log-rank test checks whether the overall descent patterns are genuinely different or just look different because of random terrain variations.
How It Works (Conceptually)
The log-rank test examines each time point where at least one event occurs across your groups. At each of these times, it:
- Counts how many events actually occurred in each group
- Calculates how many events we'd expect if there were no real difference between groups (based on the risk set sizes)
- Accumulates these observed-vs-expected differences across all event times
If the accumulated differences are large, you get a small p-value, suggesting the survival experience between groups is statistically different.
What You Get
The test produces a test statistic (usually chi-squared distributed) and a p-value. A small p-value (typically < 0.05) means you have evidence that at least one group's survival pattern differs significantly from the others.
Key Takeaway: The log-rank test is your go-to statistical tool for determining whether observed differences between survival curves are statistically significant rather than due to random chance.