scientific article; zbMATH DE number 1409846
Let \(\xi^{(n)}=(\xi_1,\xi_2,\ldots,\xi_{n})\) be observations of a Galton-Watson process with immigration \(\xi_1,\xi_2,\ldots\), where \(\xi_0=1\), \(\xi_{k}=\sum\limits _{i=1}^{\xi_{k-1}}\eta_{k,i}+1.\) Here \(\eta_{k,i}\) are i.i.d. random Poisson variables with parameter \(\theta\in (0,\infty).\) The authors consider two simple hypotheses \(H\) and \(H_1\) according to which the distribution of the observations \(\xi^{(n)}\) is defined by the measure \(P_{\theta}^{(n)}\) or \(P_{\theta_1}^{(n)}\), \(\theta\not=\theta_1.\) The authors introduce the Hellinger integral \(H_{n}(\varepsilon)\) of order \(\varepsilon\) for the measures \(P_{\theta}^{(n)}\) and \(P_{\theta_1}^{(n)}\) by setting \[ H_{n}(\varepsilon)= H(\varepsilon;P_{\theta}^{(n)},P_{\theta_1}^{(n)})= Ez_{n}^{\varepsilon},\quad \text{where}\quad z_{n}(x_1,\ldots,x_{n})= p_{\theta}(x_1,\ldots,x_{n})/ p_{\theta_1}(x_1,\ldots,x_{n}), \] \(E\) is the expectation under the hypothesis \(H\). The representation \(H_{n}(\varepsilon)=\exp\left\{ -\sum\limits_{k=1}^{n}\lambda_{k}\right\}\) is given, where \(\lambda_1,\lambda_2,\ldots\) are defined by the recurrence formula \[ \lambda_0=0, \lambda_{k}=f(\lambda_{k-1}),\quad k=1,2,\ldots,n,\quad f(\lambda)=\varepsilon\theta+(1-\varepsilon)\theta-\theta^{\varepsilon} \theta^{1-\varepsilon}e^{-\lambda}. \] They prove limit theorems for the Hellinger integral of order \(\varepsilon.\) These limit theorems are applied to prove large deviation theorems for the logarithm of the likelihood ratio. The authors investigate the rate of decrease of the 1st and 2nd type error probabilities for the Neyman--Pearson test. Note that the paper contains some misprints. Fore example, index 1 in the parameter \(\theta_1\) is missed.
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 Q4940256)
