Chernoff Efficiency of Linear Rank Statistics.
Abstract
A theorem of Hoadley used to compute large deviation probabilities for linear rank statistics with bounded score functions is extended to cover the unbounded case. The theorem is then applied to compute large deviation probabilities under alternatives in order to obtain the Chernoff efficiency of linear rank statistics. Numerical values are obtained under normal location alternatives for the two sample Wilcoxon and are shown to decrease more slowly than do corresponding Bahadur efficiency values with increasing location difference. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1972
- Accession Number
- AD0757025
Entities
People
- Tea-yuan Hwang
Organizations
- University of Wisconsin–Madison