Nonparametric Likelihood Methods for Dynamic Econometric Models: Theory and Application

本研究将探讨参数似然方法。参数似然方法是计量经济学和统计学研究的核心。一般来说，参数方法涉及在数据生成过程中使用相当任意的假设。本研究将以非参数似然方法的最新文献为基础，将似然的概念扩展到没有分布假设的模型。本研究的贡献之一是通过非参数化处理依赖结构，将各种经验似然方法推广到时间序列观测中。将进行广泛的模拟研究，以检验这些新方法的性质。此外，该项目将研究模型比较的新标准，研究动态资产定价模型的各种规格。

Yuichi Kitamura University of Minnesota - Twin Cities SBR-9632101 This research will investigate parametric likelihood methods. Parametric likelihood methods are central to econometric and statistical research. In general, parametric methods involve the use of rather arbitrary assumptions on the data generating process. This research will build on recent literature on nonparametric likelihood methods which extends the concept of likelihood to models without distributional assumptions. One of the contributions of this research is to extend various empirical likelihood methods to time series observations by treating the dependence structure nonparametrically. Extensive simulation studies will be undertaken to examine the nature of these new methods. In addition, the project will investigate a new criterion for model comparison, looking at various specifications of dynamic asset pricing models.