APPROXIMATIONS OF BEST LINEAR, UNBIASED ORDER-STATISTIC ESTIMATORS.
Abstract
Formulas for two types of approximations of the best linear, unbiased order-statistic estimator, both dispensing with the covariance matrices, have been developed and applied to the y- and z-estimators. Extensive tables for the coefficients of these estimators, their variances and covariances have been accomplished by use of an IBM 7090 computer. The efficiencies of the approximations in relation to the exact solutions have been determined for various sample sizes and shape parameters. A measure of goodness of fit has been proposed, and its mean, percentiles and variance has been determined by use of a Monte-Carlo study, involving ten thousand random samples for each of thirteen sample sizes from N = 3 to 100. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1967
- Accession Number
- AD0658845