A Regression Design Approach to Optimal and Robust Spacing Selection.
Abstract
The problem of location and/or scale parameter estimation using the asymptotically best linear unbiased estimator based on sample quantiles is considered. The problem of optimal spacing selection for these estimators is shown to be equivalent to the problem of regression design for time series with Brownian motion or Brownian bridge covariance structures and a particular variable knot spline approximation problem. This equivalence is employed, in conjunction with a regression framework, to investigate the asymptotic properties of certain spacing selection schemes. In particular, an asymptotic alternative is developed to a robust estimation procedure suggested by Chan ad Rhodin (1980). (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1981
- Accession Number
- ADA104728
Entities
People
- Randall L. Eubank
Organizations
- Southern Methodist University