Comparison of Translation Experiments

Offprint from The Annals of Mathematical Statistics, 1972, Vol. 43, No. 5, pp. 1383-1399. Original blue printed wrappers, stapled as issued; wrappers slightly worn, contents clean. Received July 29, 1970; revised September 20, 1971. Author affiliated with the University of California, Berkeley and the University of Oslo. A publisher's offprint of a significant paper in the theory of statistical experiments, treating the comparison of translation experiments within LeCam's framework of ε-deficiency; the paper generalises Boll's (1955) convolution divisibility criterion for "being more informative" to an "ε-convolution divisibility" criterion for ε-deficiency, and further generalises Strassen's (1965) convolution divisibility criterion to one for ε-convolution divisibility, showing how least favourable "ε-factors" can be found and how the deficiencies contained in an increase of information may be calculated; applications are given to rectangular, exponential, and multivariate normal experiments, determining the additional number of observations needed to compensate for deficiency, and the paper concludes by establishing that convergence for LeCam's (1964) pseudo-distance between experiments is topologically equivalent to strong convergence of the individual probability measures up to a shift.…