A TIME SERIES APPLICATION OF THE USE OF MONTE CARLO METHODS TO COMPARE STATISTICAL TESTS

D. S. Poskitt, A. R. Tremayne

Research output: Contribution to journalArticleResearchpeer-review

3 Citations (Scopus)

Abstract

Abstract. In this paper, we investigate the use of asymptotic theory to design and control simulation experiments concerning the finite sample properties of test statistics. The tests considered in the Monte Carlo experiments are various diagnostic checks which have been proposed for assessing the adequacy of the specification of open‐loop transfer function models. Attention is focussed on both the empirical size and power of the testing procedures examined and a simple descriptive measure of the relative performance of alternative tests based on changes in power resulting from variations in significance level is suggested. Whilst asymptotic theoretic considerations determine the parameter values selected, it is seen that useful information as to the actual small sample behaviour of the tests can be obtained by employing a direct modification to the standard theory.

Original languageEnglish
Pages (from-to)263-277
Number of pages15
JournalJournal of Time Series Analysis
Volume2
Issue number4
DOIs
Publication statusPublished - 1 Jan 1981
Externally publishedYes

Keywords

  • Analysis of simulations
  • Diagnostic testing
  • Experimental design
  • Local alternative
  • Transfer function model

Cite this