Minitab calculates Anderson-Darling statistics and the p-value for each distribution and transformation. For the Anderson-Darling test, the hypotheses are the following:
- H0: Data follow the distribution.
- H1: Data do not follow the distribution.
Minitab performs a likelihood-ratio test to determine whether adding an additional parameter to selected distributions significantly improves the distribution fit. For the likelihood-ratio test, the hypotheses are the following:
- H0: The simpler model (smaller distribution family) fits the data well. (Adding an extra parameter does not significantly improve the fit.)
- H1: The more complex model with an extra parameter (larger distribution family) significantly improves the fit.