Robust tests for time series with an application to first-order autoregressive processes (Q3740860)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3976152
Language Label Description Also known as
default for all languages
No label defined
    English
    Robust tests for time series with an application to first-order autoregressive processes
    scientific article; zbMATH DE number 3976152

      Statements

      Robust tests for time series with an application to first-order autoregressive processes (English)
      0 references
      0 references
      1985
      0 references
      first-order autoregressive processes
      0 references
      Rao score statistics
      0 references
      empirical power comparisons
      0 references
      Wald statistics
      0 references
      robust M-estimators
      0 references
      modified maximum likelihood equations
      0 references
      time series
      0 references
      quadratic forms
      0 references
      asymptotic relative efficiencies
      0 references
      local alternatives
      0 references
      estimating equation
      0 references
      robustness
      0 references
      0 references
      0 references
      0 references
      Let a time series model be given with parameter \(\theta\) partitioned as \((\theta_ 1,\theta_ 2)\). For estimation one-step approximations \({\tilde \theta}_ n\) to the solution of an estimating equation \(S_ n(x_ 1,...,x_ n, \theta_ n)=0\) with \(S_ n\) bounded, are considered. For testing \(\theta_ 1=\theta^ 0_ 1\) two quadratic forms based on \({\tilde \theta}_{1,n}-\theta^ 0_ 1\) and \(S_{1,n}(\theta^ 0_ 1,{\tilde \theta}_{2,n})\) respectively are proposed as test statistics. They are asymptotically chi-squared, and asymptotic relative efficiencies are calculated under local alternatives. For the estimating equation it is recommended to modify the maximum likelihood equations by truncating the estimated innovations and replacing the observations by the infinite moving average of the truncated innovations. The AR(1)-model is discussed in detail, including some simulations. Since the paper does not introduce a measure of robustness, it remains vague in which sense these tests are robust, and efficiency comparisons with other estimating equations are somewhat arbitrary.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references