A Biweight Approach to the One-Sample Problem.

Abstract

A 't'-like statistic, replacing the classical mean by a biweight in the numerator and the sample variance by a corresponding variance term in the denominator that modifies that used by Gross is proposed and evaluated for its efficiency in constructing confidence intervals in stretched-tailed situations. For the case of many samples of common population width, the method of borrowing estimates of the variance of the numerator via root-mean-squares is considered. The one-sample bi-weight-'t' is shown, via Monte Carlo simulations, to be very efficient for samples of moderate sizes (in terms of expected length of the confidence interval), and combining variance estimates in the denominator is useful when the underlying distribution of the sample is not extremely long-tailed. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 1979
Accession Number
ADA084817

Entities

People

  • Karen Kafadar

Organizations

  • Princeton University

Tags

DTIC Thesaurus Topics

  • Computing-Related Activities
  • Data Analysis
  • Data Science
  • Estimators
  • Information Science
  • Intact Stability
  • Interdisciplinary Science
  • Intervals
  • Mathematics
  • Monte Carlo Method
  • Plastic Explosives
  • Security
  • Simulations
  • Statistical Tests
  • Statistics
  • Universities

Fields of Study

  • Mathematics

Readers

  • Economics
  • Industrial Economics
  • Statistical inference.