The Kolmogorov–Smirnov test compares distributions using the largest gap D between cumulative distribution functions. Three versions are available here:
- Lilliefors: tests normality when the mean and SD are estimated from the data, with a simulated p-value.
- One sample: tests against a fully specified distribution (normal, uniform or exponential), with an exact p-value.
- Two samples: tests whether two samples come from the same distribution, with an exact p-value for small and moderate samples.
A common mistake is to run the one-sample K-S test against a normal distribution whose mean and SD were estimated from the same data. That makes the p-value far too large. Use the Lilliefors mode instead.