CALCULUS OF STATISTICAL EFFICIENCY IN A GENERAL SETTING : KERNEL PLUG-IN ESTIMATION FOR MARKOV CHAINS : HIDDEN MARKOV MODELING OF FREEWAY TRAFFIC

In the first part of this dissertation, the author develops a theoretical framework for asymptotic efficiency of statistics in non- i.i.d. cases. In the second part, the author considers the problem of estimation in Markov chains using kernel plug-in estimators. In the third part, the problem of modeling high dimensional vector time series data of freeway traffic flow collected from inductive loop detectors installed along a freeway is consider. It is assumed that an invisible binary state variable governs traffic flow at each location whose state stochastically depends on the state of neighboring locations in the previous time slice. In order to fit this hidden Markov mode, three approaches are studied: iterated conditional mode sequential importance sampling, and a modification of the model to allow a change at most one location at each time slice. Each approach is applied to data collected from the I-880 freeway, and the characteristics and behavior of each approach are studied.

  • Supplemental Notes:
    • Publication Date: 2000. UMI Company, Ann Arbor MI. Remarks: Thesis (Ph. D.)--University of California, Berkeley, 2000. Abstract also in: Dissertation abstracts international. B. Vol. 62 no. 1 (July 2001), p. 320. Format: website
  • Corporate Authors:

    University of California, Berkeley

    Department of Mechanical Engineering
    Berkeley, CA  United States  94720-1740
  • Authors:
    • Kwon, Jaimyoung
  • Publication Date: 2000

Language

  • English

Media Info

  • Pagination: 87 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00962347
  • Record Type: Publication
  • Source Agency: UC Berkeley Transportation Library
  • Report/Paper Numbers: AAT 3001909 (UMI order #)
  • Files: PATH
  • Created Date: Sep 2 2003 12:00AM