Log-rank test

The log-rank test (Mantel-Cox) checks whether the survival curves of several groups are the same. At every event time it compares the observed events per group with those expected if all groups had the same curve, and combines the deviations into a χ² with groups minus 1 degrees of freedom. With more than two groups Statable also shows pairwise tests with Holm correction.

Example: χ²(1) = 3.40, p = .065: there is no clear evidence that the curves differ.

Caution: The test is most powerful when the curves never change sides. If they cross, it can miss a real difference. It says nothing about the size of the difference; medians and proportions at fixed times serve that purpose.