Published online by Cambridge University Press: 24 October 2008
The notion of fiducial probability was introduced by R. A. Fisher and is now widely used in statistical work involving a single unknown parameter. Fisher has also considered the joint fiducial distribution of several parameters for which there exists a sufficient set of statistics, and has derived the fiducial distribution of the two parameters of the one-variate normal law*. Since Fisher's account is brief, we give a more extended description of the fiducial distribution of several independent parameters possessing a sufficient set of statistics.
* Fisher, R. A., “The fiducial argument in statistical inference”, Annals of Eugenics, 6 (1935), 391.CrossRefGoogle Scholar
* It is not difficult to show that any function of the u's and a's which has a distribution independent of the a's must be a function of the u's alone.
* Fisher, R. A., “Two new properties of mathematical likelihood”, Proc. Roy. Soc. 144 (1934), 285.CrossRefGoogle Scholar
† Bartlett, M. S., “The information available in small samples”, Proc. Cambridge Phil. Soc. 32 (1936), 560.CrossRefGoogle Scholar
‡ See Eisenhart, , Continuous groups (Princeton, 1933), p. 9.Google Scholar
* Wishart, J. and Bartlett, M. S., “The generalized product moment distribution in a normal system”, Proc. Cambridge Phil. Soc. 29 (1933), 260.CrossRefGoogle Scholar