Null Hypothesis in the Log-Rank Test
What you'll learn: You'll understand what the null hypothesis means when testing whether survival curves differ between groups.
The Core Idea
When you compare survival curves across groups (say, patients on Treatment A vs. Treatment B), you need a way to test whether any observed differences are real or just due to random chance. The null hypothesis in the Log-Rank Test provides this baseline assumption.
The null hypothesis states: The survival distributions are identical across all groups being compared.
In plain language: if the null hypothesis is true, both groups experience events (like death, churn, or machine failure) at the same rate over time. Any differences you see in your Kaplan-Meier curves are just noise—not meaningful patterns.
Why This Matters
Think of it like flipping two coins. If both coins are fair, you'd expect similar proportions of heads over many flips. But in any small sample, one might show more heads purely by luck. The null hypothesis is your "both coins are fair" assumption—it says there's no real difference in how the groups survive over time.
When you perform the Log-Rank Test, you're asking: "Is the difference between my survival curves large enough to reject this null hypothesis?" If yes, you conclude the groups truly differ. If no, you stick with the assumption that they're essentially the same.
The Alternative
If you reject the null hypothesis, you accept that at least one group has a different survival experience—maybe one treatment keeps patients alive longer, or one customer segment churns faster.
Key Takeaway: The null hypothesis in the Log-Rank Test assumes all groups have identical survival distributions—any observed differences are due to chance alone, not real group effects.