Exploring, Confirming, and Randomization Tests
Computers & Geosciences
In the context of randomization tests, this paper discusses the roles of exploratory data analysis (EDA) and confirmatory data analysis (CDA) in geoscience research. It shows: (1) how the classical methods of statistical inference can be used in EDA with nonrandom samples of data, and (2) how much of the knowledge in the geosciences is derived from EDA. The paper gives a FORTRAN IV computer program, CLASSTEST, that performs a randomization test for a multivariate analysis of variance (MANOVA) design. CLASSTEST will be useful in goescience research apart from its use in illustrating EDA and CDA.
H.C. Romesburg. 1985. Exploring, Confirming, and Randomization Tests. Computers and Geosciences, 11(1):19-37.