Chernoff index for Cox test of separate parametric families

Abstract

The asymptotic efficiency of a generalized likelihood ratio test proposed by Cox is studied under the large deviations framework for error probabilities developed by Chernoff. In particular, two separate parametric families of hypotheses are considered (Cox, 1961, 1962). The significance level is set such that the maximal type I and type II error probabilities for the generalized likelihood ratio test decay exponentially fast with the same rate. We derive the analytic form of such a rate that is also known as the Chernoff index (Chernoff, 1952), a relative efficiency measure when there is no preference between the null and the alternative hypotheses. We further extend the analysis to approximate error probabilities when the two families are not completely separated. Discussions are provided concerning the implications of the present result on model selection.

Regularized Latent Class Analysis with Application in Cognitive Diagnosis

“…$$H_0:{\mathrm{bold-italic\theta }}^{\prime }=\mathit{\theta }\text{against}{H}_1:\Vert {\mathrm{bold-italic\theta }}^{\prime }-\mathit{\theta }\Vert \ge {\epsilon }_1,{\mathrm{bold-italic\theta }}^{\prime }\in \mathrm{\Theta },$$ whose exponential decay rate has been established in Lemma 3 in Li, Liu, and Ying (2016), that there exists a rate ρ > 0 such that $P\left({sup}_{\Vert {\mathrm{bold-italic\theta }}^{\prime }-\mathit{\theta }\Vert \ge {\epsilon }_1,{\mathrm{bold-italic\theta }}^{\prime }\in \mathrm{\Theta }}\{l({\mathrm{bold-italic\theta }}^{\prime })-l(\mathit{\theta })\}\ge C_{2}N{\lambda }_N^2\right)=e^{-(\rho +o(1))N}$. Choosing ε 2 to be positive and smaller than ρ , we conclude our proof.…”

Asymptotically Optimal Sequential Design for Rank Aggregation

“…Traditional methods, such as the ones adopted in [20,45], are based on exponential change-of-measure of the log-likelihood ratio statistics, and are not directly applicable to the ranking problem considered here. The method we use in the proof combines a mixture-type of change-of-measure method recently proposed in [1,37,39] and large deviation results for martingales.…”

Uniformly efficient simulation for extremes of Gaussian random fields

“…Moreover, the proposed method can be extended to the estimation of non-Gaussian tail probabilities. For instance, in statistical hypothesis testing with data generated independently from certain distribution with unknown parameters that are of interest, it often needs to evaluate the test power/error probabilities for a range of model parameters as the sample size increase; see [22] for an example.…”