A Consistent Estimate of a Nonparametric Scale Parameter.
Abstract
A consistent estimate is proposed for the scale parameter integral of f squared in the model Y sub i = mu sub i + e sub i, 1 < or = i < or = n, where the mu sub i are unknown location parameters and the e sub i are independent, identically distributed random errors with density function f. This parameter arises in the variance formula for rank estimates of location. The proposed estimate is based on differences of residuals Y sub i *mu sub i, where *mu sub i is an estimate of mu(i). When the mu(i) follow the structure of the general linear model, the estimate is shown to be consistent under the usual assumptions on the design matrix. The estimate does not require the symmetry of the density f.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1982
- Accession Number
- ADA115803
Entities
People
- Gerald L. Sievers
Organizations
- University of North Carolina at Chapel Hill