Empirical likelihood ratio test on quantiles under a density ratio model

Population quantiles are important parameters in many applications. Enthusiasm for the development of effective statistical inference procedures and their functions has been high past decade. In this article, we study methods when multiple samples from linked populations available. The research problems consider have a wide range For example, to evolution economic status country, economists monitor changes annual household incomes, based on survey datasets collected annually. Even with samples, routine approach would estimate different separately. Such approaches ignore fact that these share some intrinsic latent structure. Recently, researchers advocated use density ratio model (DRM) account structure developed more efficient pooled data. nonparametric empirical likelihood (EL) is subsequently employed. Interestingly, there no discussion context EL-based test (ELRT) population quantiles. We explore ELRT hypotheses concerning confidence regions under DRM. show statistic chi-square limiting distribution null hypothesis. Simulation experiments distributions approximate finite-sample well lead accurate tests regions. DRM helps improve efficiency. also give real-data example illustrate efficiency proposed method.