The storage complexity of some estimators of the median

  • Finkelstein M
  • Scheinberg S
  • 2

    Readers

    Mendeley users who have this article in their library.
  • 0

    Citations

    Citations of this article.

Abstract

We consider the computational complexity of various algorithms for finding the median of an unknown distribution from which independent random samples of any size can be drawn. We present two principal results: We prove that computation of the sample median of arbitrarily large samples requires storage of unbounded amounts of data, which is intuitively reasonable. Indeed, we prove that even if the population consists of a known, finite number K of distinct elements, the computation of the medians of arbitrarily large samples requires K - 1 counters that need unbounded capacity, and no algorithm with either fewer counters or counters with bounded capacities can succeed. In contrast, we then present an algorithm, though not a very efficient one, for computing a sequence of estimators (necessarily not sample medians) that converge almost surely to the median of any given (unknown) distribution and that requires only a small, fixed number of counters (to count certain events) and registers (each of which contains a real number to the same degree of precision as the sample elements). © 1994.

Get free article suggestions today

Mendeley saves you time finding and organizing research

Sign up here
Already have an account ?Sign in

Find this document

Authors

  • Mark Finkelstein

  • Stephen Scheinberg

Cite this document

Choose a citation style from the tabs below

Save time finding and organizing research with Mendeley

Sign up for free