Answers are grounded in this paper and cite their sources.
The problem of selecting tests to be used in nonprobabilistic binary diagnostic keys is discussed. Five selection criteria are compared and it is shown that all except a new criterion suffer from some deficiency. This criterion cannot be extended easily to cope with multi-response tests but another criterion, which behaves satisfactorily with binary tests, can be extended in some circumstances. The only criterion which can always be used with multi-response tests is least satisfactory for binary tests.
No takes yet. Share an insight, caveat, or question.
Gower et al. (1975) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: