scientific article; zbMATH DE number 3926032

From MaRDI portal
Publication:3700637





[For the entire collection see Zbl 0564.00021.] An estimator \({\tilde \theta}{}_ T\) of a real parameter \(\theta\) is said to be k-th order asymptotically median unbiased (AMU) if \[ P\{\sqrt{T}({\tilde \theta}_ T-\theta_ 0)\leq x\}=2^{-1}+o(T^{-(k- 1)/2}), \] where the probability is evaluated at the parameter value \(\theta_ 0+x/\sqrt{T}\). If the k-th order Gram-Charlier expansion of a k-th order AMU estimator \({\hat \theta}{}_ T\) of a real parameter \(\theta\) coincides with that of the likelihood ratio test of a given hypothesis, then \({\hat \theta}{}_ T\) is said to be k-th order asymptotically efficient (AE). In this paper the author proves that in ARMA models depending on a scalar unknown parameter \(\theta\), the MLE and a quasi-MLE obtained by approximating the likelihood function, are both 3rd order AE. This paper is a continuation of the author's investigation of 2nd order efficiency of estimators in Gaussian ARMA models [Ann. Stat. 11, 157-169 (1983; Zbl 0509.62086)]. The initial paper will have to be read in conjunction with the present for explanations and definitions.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3700637)