Gaussian Semi‐parametric Estimation of Fractional Cointegration (Q4828159)

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 2118721
Language Label Description Also known as
default for all languages
No label defined
    English
    Gaussian Semi‐parametric Estimation of Fractional Cointegration
    scientific article; zbMATH DE number 2118721

      Statements

      Gaussian Semi‐parametric Estimation of Fractional Cointegration (English)
      0 references
      24 November 2004
      0 references
      nonstationary time series
      0 references
      multiple time series
      0 references
      long range dependence
      0 references
      long memory
      0 references
      residual-based estimation
      0 references
      A stationary time series (random process) is called \(I(d_z)\) process with memory parameter \(d_z\) if its spectral density \(f_z(\lambda)\sim G_z \lambda^{-2d_z}\) as \(\lambda \to 0^+\) for \(d_z<1/2\), \(0<G_z<\infty\). If \(z_t\) is nonstationary but has zero mean stationary increments which are \(I(d_z-1)\), \(1/2 \leq d_z <3/2\), then \(z_t\) is also said to be an \(I(d_z)\) process. A time series vector is cointegrated if a nonlinear combination of its components has lower memory than any of them and the order of cointegration, or reduction of memory, indicates the strength of the long-run relationship linking the components of the vector. The authors analyse an approach to estimation of the memory parameter of a nonstationary fractionally cointegrated vector time series using a Gaussian semi-parametric estimator based on initial consistent estimates and possibly tapered observations. The estimates of the memory parameters can rely either on original (for stationary errors) or on differenced residuals (for nonstationary errors). Thus properties of the residual periodogram are discussed as well as the main assumptions for residual-based inference. The residual-based estimates are studied under different possibilities, distinguishing consistency and asymptotic normality, with original and differenced residuals. A procedure of joint estimation of memory parameters is proposed. The authors concentrate on a bivariate system but discuss multivariate generalizations and show the performance of the estimates with simulated and real data.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references