On Frequency Polygons and Average Shifted Histograms in Higher Dimensions

Abstract

Scott has recently studied two simple variations on the ordinary histogram, namely the frequency polygon and the average shifted histogram, and found that they are able to compete with for example kernel density estimators in performance while retaining the advantage of being conceptually and computationally simple. The present paper proposes a way of generalizing frequency polygons to d-dimensional space that performs better than Scott's generalization. Expressions for integrated mean squared error and for integrated mean absolute deviation plus integrated absolute bias are obtained for generalized frequency polygons, for average shifted histograms, and for generalized frequency polygons of average shifted histograms. These expressions are used to give guidelines for window sizes. Keywords: Frequency polygons; Average shifted historgrams; Multidimensional; Integrated mean squared error; Integrated mean absolute deviation; Integrated absolute bias.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 1986
Accession Number
ADA164680

Entities

People

  • Nils L. Hjort

Organizations

  • Stanford University

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Estimators
  • Frequency
  • Histograms

Readers

  • Approximation Theory.
  • Graph Algorithms and Convex Optimization.

Technology Areas

  • Space