Recent Results on Multi-Stage Selection Procedures.
Abstract
During the past few years, several new developments took place in the area of sequential selection procedures. The purpose of the present paper is to describe the major results and to point out some open problems and interesting questions for further research in the future. The basic goal is to find that one of k populations which is associated with the largest parameter of a given underlying family of distributions. Additionally, in the control setting, one wishes to decide whether this parameter is large enough, i.e., larger than a control value. A major topic of interest is to find procedures which are reasonably economical, i.e., which perform well without requiring too many observations. The traditional criterion, due to R.E. Bechhofer, which is to guarantee the probability of a correct selection outside of a certain indifference zone, combined with the criterion of keeping the expected total sampling amount small, constitutes the main stream of current research. On the other hand, some work has also been done in the decision theoretic approach, but due to the inherent analytical difficulties, the results are rather incomplete up to now. One promising direction of further research appears to be the construction of procedures which are not too complicated and, at the same time, are at least approximately optimum in a decision theoretic sense. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1982
- Accession Number
- ADA118374
Entities
People
- Klaus J. Miescke
Organizations
- Purdue University