Problems Occurred with Histogram and a Resolution

  • Park, Byeong Uk (Department of Computer Science and Statistics, Seoul National University) ;
  • Park, Hong Nae (Department of Computer Science and Statistics, Seoul National University) ;
  • Song, Moon Sup (Department of Computer Science and Statistics, Seoul National University) ;
  • Song, Jae Kee (Department of Statistics, Kyungbook National University)
  • Published : 1990.11.01

Abstract

In this article, several problems inherent in histogram estimate of unknown probability density function are discussed. Those include so called sharp comers and bin edge effect. A resolution for these problems occurred with histogram is discussed. The resulting estimate is called kernel density estimate which is most widely used by data analysts. One of the most recent and reliable data-based choices of scale factor (bandwidth) of the estimate, which has been known to be most crucial, is also discussed.

Keywords

Acknowledgement

Supported by : Ministry of Education