A0512
Title: Pivot statistics for normal populations
Authors: Liqiang Ni - University of Central Florida (United States) [presenting]
Abstract: Pivot statistics are widely used as the basis for statistical inference, e.g. confidence interval, hypothesis testing, and predictive inference. The concepts of sufficiency, pivot, invariance, and Bayesian inference are introduced. Then, the linkage between them is discussed, and some surprising results are shown: some classic statistic long thought to be pivot is not pivot; reducing the requirement of affine invariance to low-triangular invariance can produce a pivot. Some work-in-progress in the search for "true" pivots are considered.
