Nonparametric methods of process discrimination and model validation using zero crossings

  • Omer F. Demirel*
  • , Thomas R. Willemain
  • *Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

We address two related situations. Process discrimination. One has many sample realizations from each of two stationary data generating processes. Without making parametric models of the processes, one wants to test whether the two data generating processes are the same. Model validation. One has a single realization generated by a stationary real-world process. One also has one or perhaps a few realizations generated by a model of the process. One wants to assess the validity of the model. Both these problems involve the comparison of two sets or pairs of time series data. Our solutions use a new statistic which measures the "distance" between two time series based on their zero crossings. This article describes the statistic and several Monte Carlo studies which establish its utility for process discrimination and model validation.

Original languageEnglish
Pages (from-to)517-539
Number of pages23
JournalCommunications in Statistics Part B: Simulation and Computation
Volume32
Issue number2
DOIs
StatePublished - May 2003

Bibliographical note

Funding Information:
We acknowledge helpful conversations with Jorge Haddock, Nong Shang, Pasquale Sullo, Balaji Rajagopalan, and Upmanu Lall. We also benefited from comments from two anonymous referees. Randy Norsworthy provided the financial data. This work was supported by National Science Foundation grant DMI 9813097.

Keywords

  • Simulation
  • Time series

ASJC Scopus subject areas

  • Statistics and Probability
  • Modeling and Simulation

Fingerprint

Dive into the research topics of 'Nonparametric methods of process discrimination and model validation using zero crossings'. Together they form a unique fingerprint.

Cite this