Answers are grounded in this paper and cite their sources.
SUMMARY The structure of a data layout may be usefully described in terms of its symmetries, i.e. transformations which do not affect attitudes to the data. Interesting null hypotheses may be considered as imposing still further symmetries. These symmetry properties may be used to derive valid tests of such null hypotheses, and to construct appropriate statistical models for the data. The general ideas are developed and illustrated in the context of a replicated two-way classification.
No takes yet. Share an insight, caveat, or question.
A. P. Dawid (1988) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: