STATISTICS WITHOUT PROBABILITY - SIGNIFICANCE TESTING AS TYPICALITY AND EXCHANGEABILITY IN DATA-ANALYSIS

Statistical significance is almost universally equated with the attrib ution to some population of nonchance influences as the source of stru cture in the data. But statistical significance can be divorced from b oth parameter estimation and probability as, instead, a statement abou t the atypicality or lack of exchangeability over some distinction of the data relative to some set. From this perspective, the criticisms o f significance tests evaporate.