No label defined (Q3706330)
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 3934212
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 3934212 |
Statements
1982
0 references
consistency
0 references
higher order asymptotic efficiency
0 references
regular estimators
0 references
stochastic expansion
0 references
Edgeworth series
0 references
asymptotic mean
0 references
modified maximum likelihood estimator
0 references
concentration
0 references
quadratic form of normal variables
0 references
symmetry of the loss function
0 references
0.8024317026138306
0 references
0.7920668125152588
0 references
0.7920668125152588
0 references
[For the entire collection see Zbl 0564.00018.] Recently, higher order asymptotic efficiency of especially the maximum likelihood estimators (m. 1. e.) has been established by J. Pfanzagl et. al., J. K. Ghosh et. al., and the author and M. Akahira. Essentially it is shown that for regular estimators \(\{\) \({\hat \theta}{}^ 0_ n\}\) of a real parameter \(\theta\) for which \(\sqrt{n}({\hat \theta}^ 0_ n-\theta)\) a) has a stochastic expansion, b) has an asymptotic distribution expansible into an Edgeworth series, and c) has asymptotic mean of order \(n^{-1/2}\), the modified maximum likelihood estimator \({\hat \theta}{}^*_ n\) has the greatest possible concentration, i.e. \[ \lim_{n\to \infty}n[P\{\sqrt{n}| {\hat \theta}^*_ n-\theta | <a\}-P\{\sqrt{n}| {\hat \theta}^ 0_ n-\theta | <a\}]\geq 0 \] \[ for\quad all\quad a>0,\quad where\quad {\hat \theta}^*_ n={\hat \theta}_ n+n^{-1}g({\hat \theta}_ n) \] and \({\hat \theta}{}_ n\) is the m. 1. e. and g is a function defined corresponding to \({\hat \theta}{}^ 0_ n.\) Moreover if the class of estimators is further restricted to the class \(D^*\) in that d) the second order term in the stochastic expansion is asymptotically equivalent to a quadratic form of normal variables, we have \[ \lim_{n\to \infty}n[P\{-a<\sqrt{n}({\hat \theta}^*_ n- \theta)<b\}-P\{-a<\sqrt{n}({\hat \theta}_ n-\theta)<b\}]\geq 0 \] for any \(a>0\) and \(b>0.\) The purpose of this paper is to investigate the implications of the above results in decision-theoretic set-ups, where it will be shown that the symmetry of the loss function is the most important criterion to be accounted for.
0 references