Preferred Point Geometry and Statistical Manifolds
Annals Of Statistics, 1993, 21, 3, 1197-1224
CRITCHLEY, Frank, MARRIOTT, Paul, SALMON, Mark, Preferred Point Geometry and Statistical Manifolds, Annals Of Statistics, 1993, 21, 3, 1197-1224 - https://hdl.handle.net/1814/16763
Retrieved from Cadmus, EUI Research Repository
A new mathematical object called a preferred point geometry is introduced in order to (a) provide a natural geometric framework in which to do statistical inference and (b) reflect the distinction between homogeneous aspects (e.g., any point theta may be the true parameter) and preferred point ones (e.g., when theta0 is the true parameter). Although preferred point geometry is applicable generally in statistics, we focus here on its relationship to statistical manifolds, in particular to Amari's expected geometry. A symmetry condition characterises when a preferred point geometry both subsumes a statistical manifold and, simultaneously, generalises it to arbitrary order. There are corresponding links with Barndorff-Nielsen's strings. The rather unnatural mixing of metric and nonmetric connections in statistical manifolds is avoided since all connections used are shown to be metric. An interpretation of duality of statistical manifolds is given in terms of the relation between the score vector and the maximum likelihood estimate.
Cadmus permanent link: https://hdl.handle.net/1814/16763
Full-text via DOI: 10.1214/aos/1176349258
Earlier different version: http://hdl.handle.net/1814/390
Version: The article is a published version of EUI ECO WP; 1991/51
Files associated with this item
There are no files associated with this item.