Answers are grounded in this paper and cite their sources.
The problem of comparing means from normal distributions with standard deviations a-, and o-2 iS considered from two aspects, using on the one hand the concepts of probabilities of chance events, reference sets to which these probabilities refer, and verification (i.e. realization in long series of experiments) of the probability statements, and on the other hand the concept of the random variable. (a) If previous experience provides no information about the parameter values appropriate for a particular experiment, the fiducial argument is justified, and leads to the Fisher-Behrens test. (b) If previous experience provides adequate density functions for the variances, Bayes' theorem provides a solution to the problem. (c) For situations intermediate between (a) and (b), the Fisher-Behrens solution is approximately valid.
No takes yet. Share an insight, caveat, or question.
W.-U. Behrens (1964) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: