
A distribution comparison weighted towards the tails, where the Kolmogorov-Smirnov test is weakest.The Anderson–Darling test is a statistical test used to decide whether a sample of data comes from a specific probability distribution. It is a refinement of the older Kolmogorov–Smirnov test, designed to be more sensitive to differences in the tails of the distribution — exactly where the Kolmogorov–Smirnov test tends to miss deviations. This makes it particularly useful in fields like quality control and finance, where extreme events or outliers matter more than the central tendency.
The test was introduced by Theodore W. Anderson and Donald A. Darling in 1952.
Like other goodness-of-fit tests, it compares the empirical cumulative distribution function of the sample with the theoretical cumulative distribution function of the hypothesized distribution. The test statistic is a weighted squared difference between the two, with weights that are larger at the tails. The resulting value is compared against critical values (which depend on the hypothesized distribution) to decide whether the sample is consistent with that distribution.
It is widely used in statistical process control to check whether a production process output follows a normal distribution, and in risk management to validate models of extreme events.
The Anderson–Darling test is so sensitive to the tails that it can detect a slight heavy-tailedness in a sample that looks perfectly normal on a histogram.

