不独立序列情形下累积分布函数光滑同时置信带.pdfVIP

英文摘要 不独立序列情形下累积分布函数的光滑同时置信带 Sm ooth Sim ultaneous Confidence B ands for C um ulative D istribution Functions under D ependent Variables A bstract Literatures have developed that a plug-in kernel estimator can be used as a smooth estimator of a cumulative distribution function with uniform convergence if indepen­ dent samples given. Does this conclusion still hold in the case of dependent samples? In this paper, we propose an smooth estimator called kernel distribution estimator based on the dependent samples, and then prove the uniform convergence between this estimator and the empirical distribution function if some mild assumptions are satis­ fied. While according to the knowledge about mixing sequence in some relating paper, it has been established the uniform convergence between the cumulative distribution and the empirical distribution function under certain assumptions. So when the both condition are satisfied, we get the convergence between the plug-in kernel estimator and the cumulative distribution function. Furthermore, we construct a smooth confi­ dence band and apply this method to an extensive simulation study. In this paper, we use two different automatic bandwidths to make simulations and the simulation results confirm our theory. Keywords: Holder continuity; Kernel function; Mixing sequence; Simulation confi­ dence bands; Uniform convergence. Written by: Xia Qin Supervised by: Prof. Lijian Yan II 目 录 第 一 章 引 言 1 第二章主 要定 理及 推论 2 第三章 构 造 方法 实施 4 第四章 数据 模拟 研究 5 § 4 .1 全 局 性质的 比较 5 §4.2 同时置信带 9 第 五 章 定