Empirical processes for infinite variance autoregressive models (Q413784): Difference between revisions
From MaRDI portal
Set profile property. |
Set OpenAlex properties. |
||
Property / full work available at URL | |||
Property / full work available at URL: https://doi.org/10.1016/j.jmva.2012.01.018 / rank | |||
Normal rank | |||
Property / OpenAlex ID | |||
Property / OpenAlex ID: W3121848610 / rank | |||
Normal rank |
Revision as of 21:28, 19 March 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Empirical processes for infinite variance autoregressive models |
scientific article |
Statements
Empirical processes for infinite variance autoregressive models (English)
0 references
7 May 2012
0 references
Consider an \(\text{AR}(p)\) model \(X_i= \sum^p_{l=1} \phi_l X_{i-l}+ \varepsilon_i\), where the \(\varepsilon_i\)'s are independent and have distribution \(F\) belonging to the domain of attraction of a stable law with index \(\alpha\in(0, 2)\). Let \(\widehat\phi= (\widehat\phi_0,\dots, \widehat\phi_l)\) be a consistent estimate of \(\phi= (\phi_0,\dots,\phi_l)\) and define the residuals \(e_{in}\;,1\leq i\leq n,\) by \[ e_{in}= X_i- \widehat\phi_0- \sum^p_{l=1} \widehat\phi_l X_{i-l}\qquad (1\leq i\leq n). \] For \({\mathbf t}= (t_1,\dots, t_m)\) define \[ \begin{aligned} \widetilde{\mathbf F}_{n,m}({\mathbf t}) &=n_m^{-1} \sum^{n_m}_{i=1} I\,(\varepsilon_i\leq t_1,\dots, \varepsilon_{i+m-1}\leq t_m),\\ {\mathbf F}_m({\mathbf t}) &= \prod^m_{l=1} F(t_l),\quad \widetilde{\mathbf K}_{n,m}({\mathbf t})= \sqrt{n_m} (\widetilde{\mathbf F}_{n,m}({\mathbf t})-{\mathbf F}_m({\mathbf t})).\end{aligned} \] Assume that \(\sqrt{n}(\widehat\phi_0- \phi_0)\) converges weakly to some \(Z_0\) and that \(\widetilde a_n(\widehat\phi_1- \phi_1,\dots, \widehat\phi_p- \phi_p)\) converges weakly to \((Z_1,\dots,Z_p)\) for some sequence \(\widetilde a_n= n^{1/\alpha}\widetilde L(n)\) with \(\widetilde L\) a slowly varying function. In this paper, among others, the authors establish that under some additional conditions \(\widetilde{\mathbf K}_{n,m}({\mathbf t})\) converges weakly to some process \(\widetilde{\mathbf K}_m\) having an explicit representation. The results are applied to tests of goodness-of-fit and randomness and to the classical portmanteau type statistics. A simulation study is also given.
0 references
stable distributions
0 references
independence tests
0 references
goodness-of-fit tests
0 references
portmanteau statistics
0 references