New Approaches to Robust Confidence Intervals for Location: A Simulation Study.
Abstract
The robustness of validity of four methods for setting confidence intervals for a location parameter when the scale is unknown are investigated. Three methods involve estimating the variance of an M- estimate of location while the fourth is a procedure suggested by Maritz, based on a permutation argument. The first three methods use either a finite sample approximation to the asymptotic variance (a well-known standard) or make inferences on the basis of the shape of the putative likelihood function. The latter approach is related to the work of Sprott, as well as that of Efron and Hinkley on conditional inference. Overall, the Maritz procedure performs best though the standard does suprisingly well. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1984
- Accession Number
- ADA142248
Entities
People
- D. Pregibon
- H. I. Braun
Organizations
- Educational Testing Service