Abstract

Modern statistical science, as counted from the time of R. A. Fisher (circa 1922), is rooted in several key ideas for data reduction, summary, and interpretation. One such powerful idea is statistical sufficiency. Sufficiency has received renewed attention as data reduction and interpretation are becoming even more important in these data-rich times. In this talk I will give an overview of the evolution of this important idea and explain how it has permeated many aspects of contemporary statistics. I will focus on two areas of my own research where sufficiency has played a key role: sufficient dimension reduction and sufficient graphical models.