日時 2001年5月 8日(火) 15時〜16時40分 場所 経済学部5階視聴覚室 講演者 Christopher Small(University of Waterloo) 演題 Likelihood Tilting 概要 The concept of a tilted distribution has proved to be of great value in statistical inference. In this talk, I shall discuss some examples of tilting, and shall examine how likelihood tilting, in particular, can provide a different paradigm for likelihood methodology. The integrated tilt of a distribution will be developed as an operator, and model assumptions shall be interpreted as families of operators acting on distributions. Examples from parametric and semiparametric inference will be studied.