A New Kind of Asymptotic Quasi‐Score Estimating Function (Q4956068)

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 1452481
Language Label Description Also known as
default for all languages
No label defined
    English
    A New Kind of Asymptotic Quasi‐Score Estimating Function
    scientific article; zbMATH DE number 1452481

      Statements

      A New Kind of Asymptotic Quasi‐Score Estimating Function (English)
      0 references
      0 references
      24 May 2000
      0 references
      estimating functions
      0 references
      asymptotic quasi-likelihood
      0 references
      consistency
      0 references
      The author proposes a new definition of asymptotic quasi-score estimating functions and proves that under some conditions consistency of the asymptotic quasi-score estimator is equivalent to the consistency of a quasi-likelihood estimator. It is supposed that the data are described by the model \(y_t=f_t(\vartheta)+m_t\), where \(\vartheta\in\Theta\) is an unknown parameter, \(f_t\) is a predictable process and \(m_t\) is a martingale difference process w.r.t. the \(\sigma\)-fields \(F_t\), \(t=0,1,\dots,T\), \(0<E(m^2_t |F_{t-1})<\infty\). The estimating functions are chosen from the class \[ \Psi_T=\left\{\sum_{t=1}^T a_tm_t \;|\;a_t \text{ is } F_{t-1} \text{ measurable}\right\}. \] Denote \( {\mathcal E}(G_T)=(E_\vartheta \dot{G}_T)^{-1} E_\vartheta(G_TG'_T)(E_\vartheta \dot{G}'_T)^{-1} \). For any class \({\mathcal G}_T\subseteq\Psi_T\) such that \(\forall G_T\in{\mathcal G}_T\) holds \(E_\vartheta G(\vartheta)=0\) the quasi-score function is defined as \(G^*_T\) for which \({\mathcal E}(C_T)-{\mathcal E}(C^*_T)\) is nonnegative definite. The asymptotic quasi-score is a sequence \(G_{T,n}^*\) in \({\mathcal G}_T\) such that \({\mathcal E}(C_T)-{\mathcal E}(C^*_{T,n})\) is asymptotically nonnegative definite. For the considered model \[ G^*_T=\sum_{t=0}^T(f_t(\vartheta)m_t/E(m^2_t |F_{t-1}) \] but in many applications it depends on nuisance parameters and can not be calculated. The author proposes a practical procedure for asymptotic quasi-score construction and considers some simulation results.
      0 references

      Identifiers