Research on Statistics

Abstract

Since its founding in 1947, ETS has supported research in a variety of areas-a fact attested to by the many different chapters comprising this volume. As a private, nonprofit organization known primarily for its products and services related to standardized testing, it comes as no surprise that ETS conducted extensive research in educational measurement and psychometrics, which together provide the scientific foundations for the testing industry. This work is documented in the chapters in this book. At the same time, a good part of educational measurement and perhaps most of psychometrics can be thought of as drawing upon-and providing an impetus for extending-work in theoretical and applied statistics. Indeed, many important developments in statistics are to be found in the reports alluded to above.One may ask, therefore, if there is a need for a separate chapter on statistics. The short answer is yes. The long answer can be found in the rest of the chapter. A review of the ETS Research Report (RR) series and other archival materials reveals that a great deal of research in both theoretical and applied statistics was carried out at ETS, both by regular staff members and by visitors. Some of the research was motivated by longstanding problems in statistics, such as the Behrens-Fisher problem or the problem of simultaneous inference, and some by issues arising at ETS during the course of business. Much of this work is distinguished by both its depth and generality. Although a good deal of statistics-related research is treated in other chapters, much is not.The purpose of this chapter, then, is to tell a story of statistics research at ETS. It is not the story, as it is not complete; rather, it is structured in terms of a number of major domains and, within each domain, a roughly chronological narrative of key highlights. As will be evident, the boundaries between domains are semipermeable so that the various narratives sometimes intermix. Consequently, reference will also